#  @_reachsumit Sumit Sumit posts on X about kuaishou, generative, alibaba, llm the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours. ### Engagements: [-----] [#](/creator/twitter::129969370/interactions)  - [--] Week [------] +62% - [--] Month [------] +17% - [--] Months [-------] +166% - [--] Year [-------] -54% ### Mentions: [--] [#](/creator/twitter::129969370/posts_active)  - [--] Month [--] -67% - [--] Months [---] +62% - [--] Year [---] +81% ### Followers: [-----] [#](/creator/twitter::129969370/followers)  - [--] Week [-----] +0.74% - [--] Month [-----] +3% - [--] Months [-----] +14% - [--] Year [-----] +29% ### CreatorRank: [---------] [#](/creator/twitter::129969370/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) [social networks](/list/social-networks) [stocks](/list/stocks) [finance](/list/finance) [fashion brands](/list/fashion-brands) [currencies](/list/currencies) [cryptocurrencies](/list/cryptocurrencies) **Social topic influence** [kuaishou](/topic/kuaishou) #53, [generative](/topic/generative) #280, [alibaba](/topic/alibaba) #361, [llm](/topic/llm) #408, [tencent](/topic/tencent) #175, [bytedance](/topic/bytedance), [meituan](/topic/meituan) #16, [prediction](/topic/prediction), [meta](/topic/meta), [realtime](/topic/realtime) **Top accounts mentioned or mentioned by** [@jeremyphoward](/creator/undefined) [@bowenjin13](/creator/undefined) [@xiaopengli](/creator/undefined) [@kvachai](/creator/undefined) [@youganglyu](/creator/undefined) [@junruwu4](/creator/undefined) [@zeyuanmeng](/creator/undefined) [@louisackerman74](/creator/undefined) [@genksn4](/creator/undefined) [@yashardel](/creator/undefined) [@rayluthu](/creator/undefined) [@jinyeopsong](/creator/undefined) [@seungheondoh](/creator/undefined) [@seirasto](/creator/undefined) [@gsehgal1997](/creator/undefined) [@guz](/creator/undefined) [@hschechtr](/creator/undefined) [@mufeili](/creator/undefined) [@jlwu002](/creator/undefined) [@aksh555](/creator/undefined) **Top assets mentioned** [Microsoft Corp. (MSFT)](/topic/microsoft) [Alphabet Inc Class A (GOOGL)](/topic/$googl) [IBM (IBM)](/topic/ibm) [Spotify Technology (SPOT)](/topic/$spot) [Snap, Inc. (SNAP)](/topic/snap-inc) [Lossless (LSS)](/topic/lossless) [Intuit Inc. (INTU)](/topic/intuit) [Merge (MERGE)](/topic/merge) [Frontier (FRONT)](/topic/frontier) [Fusion (FSN)](/topic/fusion) [Linear (LINA)](/topic/linear) [Fetch (FET)](/topic/fetch) ### Top Social Posts Top posts by engagements in the last [--] hours "Revisiting Neural Retrieval on Accelerators Proposes a non-dot-product retrieval approach called "mixture of logits" (MoL) that models user-item similarity as an adaptive composition of elementary similarity functions. https://arxiv.org/abs/2306.04039 https://arxiv.org/abs/2306.04039" [X Link](https://x.com/_reachsumit/status/1666690781367898112) 2023-06-08T06:17Z [----] followers, [---] engagements "Contrastive Multi-view Framework for Customer Lifetime Value Prediction Utilizes multiple heterogeneous regressors to extract complementary knowledge synthesizes predictions into a unified score & combines it with a purchase classifier to predict LTV https://arxiv.org/abs/2306.14400 https://arxiv.org/abs/2306.14400" [X Link](https://x.com/_reachsumit/status/1673539569722216449) 2023-06-27T03:51Z [----] followers, [--] engagements "Structure Guided Multi-modal Pre-trained Transformer for Knowledge Graph Reasoning Leverages the structural information in multimodal knowledge graphs by adopting a graph structure encoder and a structure-guided fusion module" [X Link](https://x.com/_reachsumit/status/1678238515208032257) 2023-07-10T03:13Z [---] followers, [---] engagements "Lexically-Accelerated Dense Retrieval Proposes a technique using lexical retrieval to seed and accelerate dense retrieval exploration via a document proximity graph. https://github.com/Georgetown-IR-Lab/ladr https://arxiv.org/abs/2307.16779 https://github.com/Georgetown-IR-Lab/ladr https://arxiv.org/abs/2307.16779" [X Link](https://x.com/_reachsumit/status/1686384002461208576) 2023-08-01T14:30Z [----] followers, [----] engagements "NEON: Living Needs Prediction System in Meituan Proposes a system to predict living needs using spatiotemporal and behavioral features with multitask learning deployed and evaluated on Meituan" [X Link](https://x.com/_reachsumit/status/1686385854229704704) 2023-08-01T14:38Z [---] followers, [---] engagements "SPM: Structured Pretraining and Matching Architectures for Relevance Modeling in Meituan Search Proposes a relevance matching architecture for ecommerce search queries and structured documents using multiple document fields and query-guided extractors" [X Link](https://x.com/_reachsumit/status/1691675426845053243) 2023-08-16T04:57Z [---] followers, [---] engagements "I published Vol. [--] of "Top Information Retrieval Papers of the Week" on Substack" [X Link](https://x.com/_reachsumit/status/1693268246047375484) 2023-08-20T14:26Z [---] followers, [--] engagements "Exploring the Integration Strategies of Retriever and Large Language Models Explores ways to improve open-domain question answering by better integrating retrieved passages with large language models and suggests four strategies. https://arxiv.org/abs/2308.12574 https://arxiv.org/abs/2308.12574" [X Link](https://x.com/_reachsumit/status/1694923321799000081) 2023-08-25T04:03Z [----] followers, [---] engagements "Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations Combines recommender models and LLMs for versatile and interactive recommendation through efficient components like memory bus and demonstration-augmented planning" [X Link](https://x.com/_reachsumit/status/1697432115795857541) 2023-09-01T02:12Z [---] followers, [--] engagements "Personalized Search Via Neural Contextual Semantic Relevance Ranking Proposes a personalized LTR framework that models relationships between user query context and document content using lexical and semantic representations to improve search relevance" [X Link](https://x.com/_reachsumit/status/1701438578356801583) 2023-09-12T03:32Z [---] followers, [---] engagements "Retrieving Supporting Evidence for Generative Question Answering Explores using large language models to self-detect hallucinations in generated text by prompting the model to verify its responses against supporting evidence retrieved from a corpus" [X Link](https://x.com/_reachsumit/status/1704728107289399550) 2023-09-21T05:23Z [---] followers, [--] engagements "Cluster Language Model for Improved E-Commerce Retrieval and Ranking: Leveraging Query Similarity and Fine-Tuning for Personalized Results Combines pre-trained LMs with query clustering to offer improved e-commerce retrieval and ranking performance" [X Link](https://x.com/_reachsumit/status/1706760680215544028) 2023-09-26T20:00Z [---] followers, [---] engagements "RankVicuna: Zero-Shot Listwise Document Reranking with Open-Source Large Language Models Introduces an open-source listwise reranking approach using a 7B parameter language model for zero-shot document reranking" [X Link](https://x.com/_reachsumit/status/1706891397721366562) 2023-09-27T04:39Z [---] followers, [---] engagements "The optimal way to utilize multimodal features is through end-to-end training of the recommender model and the video encoder. This is likely because E2E training can incorporate both raw video features and collaborative signals from user-item interactions" [X Link](https://x.com/_reachsumit/status/1707260857682739377) 2023-09-28T05:08Z [---] followers, [--] engagements "Aligning the Capabilities of Large Language Models with the Context of Information Retrieval via Contrastive Feedback Proposes a framework called RLCF to align LLMs to improve their ability to generate specific and distinguishable responses" [X Link](https://x.com/_reachsumit/status/1708667456615748033) 2023-10-02T02:17Z [---] followers, [----] engagements "Our Modern Recommendation Systems (MRS) team at Meta is hiring IC4/5/6 MLEs. Join us" [X Link](https://x.com/_reachsumit/status/1710455474519466465) 2023-10-07T00:42Z [---] followers, [----] engagements "Lending Interaction Wings to Recommender Systems with Conversational Agents Proposes a plug-and-play framework that incorporates conversational agents into recsys by treating the recommender as a relevance estimator and the agent as a score checker" [X Link](https://x.com/_reachsumit/status/1711543121824395429) 2023-10-10T00:44Z [---] followers, [---] engagements "Factual and Personalized Recommendations using Language Models and Reinforcement Learning Proposes a "personalized and compelling" language model for conversational recommender systems which uses reinforcement learning with AI feedback" [X Link](https://x.com/_reachsumit/status/1711986671485661251) 2023-10-11T06:06Z [---] followers, [---] engagements "Retrieve Anything To Augment Large Language Models Proposes a unified embedding model trained to support diverse retrieval needs of large language models including knowledge memory examples and tools" [X Link](https://x.com/_reachsumit/status/1712285198963048878) 2023-10-12T01:52Z [---] followers, [---] engagements "Found in the Middle: Permutation Self-Consistency Improves Listwise Ranking in Large Language Models Proposes generating multiple rankings from permutations of a prompt list and aggregates them to mitigate positional biases" [X Link](https://x.com/_reachsumit/status/1712306343330328790) 2023-10-12T03:17Z [---] followers, [---] engagements "A Setwise Approach for Effective and Highly Efficient Zero-shot Ranking with Large Language Models Proposes a prompting approach for zero-shot document ranking with LLMs outperforming existing methods" [X Link](https://x.com/_reachsumit/status/1714511893476639138) 2023-10-18T05:21Z [---] followers, [---] engagements "Chameleon: a Heterogeneous and Disaggregated Accelerator System for Retrieval-Augmented Language Models Proposes an accelerator system with specialized hardware for language model inference & vector search to efficiently serve diverse RALM configs" [X Link](https://x.com/_reachsumit/status/1714514704327119315) 2023-10-18T05:32Z [---] followers, [---] engagements "Search-Adaptor: Text Embedding Customization for Information Retrieval Google Cloud AI proposes an adaptation method that improves text embeddings from LLMs (even API-based ones) for information retrieval by using a ranking loss and regularizers" [X Link](https://x.com/_reachsumit/status/1714517186340118669) 2023-10-18T05:42Z [---] followers, [----] engagements "A Comprehensive Survey on Vector Database: Storage and Retrieval Technique Challenge Provides a review of vector database architectures approximate nearest neighbor search algorithms challenges and emerging directions involving LLMs" [X Link](https://x.com/_reachsumit/status/1714852244028571987) 2023-10-19T03:53Z [---] followers, [---] engagements "Self-RAG: Learning to Retrieve Generate and Critique through Self-Reflection Proposes a framework that trains language models to retrieve knowledge on-demand and reflect on retrieved passages and their own generations" [X Link](https://x.com/_reachsumit/status/1714854420553568738) 2023-10-19T04:02Z [---] followers, [---] engagements "Know Where to Go: Make LLM a Relevant Responsible and Trustworthy Searcher Proposes a generative retrieval framework with generator validator optimizer modules to improve the relevance responsibility & trustfulness of LLM-based retrieval systems" [X Link](https://x.com/_reachsumit/status/1715377521011859902) 2023-10-20T14:40Z [---] followers, [---] engagements "Large Search Model: Redefining Search Stack in the Era of LLMs Microsoft proposes using a single large language model for all search tasks instead of many specialized models formulating tasks as text generation from prompts" [X Link](https://x.com/_reachsumit/status/1717618675295904178) 2023-10-26T19:06Z [---] followers, 13.8K engagements "PaRaDe: Passage Ranking using Demonstrations with Large Language Models Google explores using difficulty-based selection to automatically choose effective demonstrations for improving LLM passage ranking and question generation" [X Link](https://x.com/_reachsumit/status/1717620214018580562) 2023-10-26T19:12Z [---] followers, [---] engagements "One Model for All: Large Language Models are Domain-Agnostic Recommendation Systems Ant Group proposes using LLMs for multi-domain sequential recommendation by representing users and items with concatenated item titles to leverage pretrained knowledge" [X Link](https://x.com/_reachsumit/status/1717622399607144949) 2023-10-26T19:21Z [---] followers, [----] engagements "Representation Learning with Large Language Models for Recommendation Proposes a framework to enhance existing recommender systems by aligning semantic representations from LLMs with collaborative representations" [X Link](https://x.com/_reachsumit/status/1717629434390274539) 2023-10-26T19:49Z [---] followers, [---] engagements "Open-source Large Language Models are Strong Zero-shot Query Likelihood Models for Document Ranking Investigates the zero-shot ranking effectiveness of LLMs as QLMs and proposes to integrate them with a hybrid retriever" [X Link](https://x.com/_reachsumit/status/1717767518276743515) 2023-10-27T04:57Z [---] followers, [---] engagements "CoLLM: Integrating Collaborative Embeddings into Large Language Models for Recommendation Proposes to enhance LLM recommenders by incorporating collaborative information from traditional models thru embedding space alignment" [X Link](https://x.com/_reachsumit/status/1719526470593425669) 2023-11-01T01:27Z [---] followers, [---] engagements "MILL: Mutual Verification with Large Language Models for Zero-Shot Query Expansion Proposes a framework for query expansion which generates documents through q-q-d prompts and mutually filters retrieved and generated documents to complement each other" [X Link](https://x.com/_reachsumit/status/1719533178078265462) 2023-11-01T01:53Z [---] followers, [----] engagements "LLMRec: Large Language Models with Graph Augmentation for Recommendation Proposes a framework that LLMs to augment recommender systems by reinforcing user-item edges enhancing item attributes and profiling users" [X Link](https://x.com/_reachsumit/status/1719900318799163687) 2023-11-02T02:12Z [---] followers, [----] engagements "Collaborative Large Language Model for Recommender Systems LinkedIn proposes a generative recommender system that tightly integrates the ID paradigm and LLM paradigm using extended vocabulary novel pretraining and finetuning" [X Link](https://x.com/_reachsumit/status/1720450366893129967) 2023-11-03T14:38Z [---] followers, [----] engagements "Instruction Distillation Makes Large Language Models Efficient Zero-shot Rankers Distills complex pairwise ranking instructions into simpler pointwise instructions to improve the effectiveness of LLMs for zero-shot ranking" [X Link](https://x.com/_reachsumit/status/1721412667796300211) 2023-11-06T06:22Z [---] followers, 10.5K engagements "LlamaRec: Two-Stage Recommendation using Large Language Models for Ranking Proposes a two-stage recommendation framework that uses a sequential model for efficient candidate retrieval and Llama [--] for effective ranking" [X Link](https://x.com/_reachsumit/status/1721741092515938348) 2023-11-07T04:07Z [---] followers, [----] engagements "Knowledge-Augmented Large Language Models for Personalized Contextual Query Suggestion Microsoft Research presents a method to personalize LLMs for search via entity-based user knowledge stores derived from logs" [X Link](https://x.com/_reachsumit/status/1724322136372977765) 2023-11-14T07:03Z [---] followers, 19.1K engagements "Text Retrieval with Multi-Stage Re-Ranking Models Hitachi roposes a 3-stage retrieval model using BM25 language model and model ensemble/larger model to improve accuracy while reducing search delays" [X Link](https://x.com/_reachsumit/status/1724686264480477444) 2023-11-15T07:10Z [---] followers, [---] engagements "Chain-of-Note: Enhancing Robustness in Retrieval-Augmented Language Models Tencent AI Lab proposes Chain-of-Noting to improve retrieval-augmented language models through robustness to irrelevant retrieved documents and unknown queries" [X Link](https://x.com/_reachsumit/status/1725032167171043791) 2023-11-16T06:04Z [---] followers, [---] engagements "@jeremyphoward Congratulations Pat Cummins was absolutely brilliant tonight and a nightmare to face as the opposing team's captain. I'll hold a grudge against him for the rest of his cricketing career. ðŸ˜ðŸ˜ðŸ˜‚" [X Link](https://x.com/_reachsumit/status/1726362492727930956) 2023-11-19T22:11Z [---] followers, [---] engagements "RankingGPT: Empowering Large Language Models in Text Ranking with Progressive Enhancement Alibaba trains LLMs for text ranking through weak supervision & fine-tuning to align objectives to achieve state-of-the-art performance" [X Link](https://x.com/_reachsumit/status/1729710949047607555) 2023-11-29T03:56Z [---] followers, [---] engagements "ControlRec: Bridging the Semantic Gap between Language Model and Personalized Recommendation Meituan proposes to align user/item IDs with language semantics in recommendations via contrastive learning for improved generalization" [X Link](https://x.com/_reachsumit/status/1729712604321370372) 2023-11-29T04:03Z [---] followers, [---] engagements "Event-driven Real-time Retrieval in Web Search Tencent proposes Event-driven Real-time Retrieval an approach that expands queries with breaking events to improve real-time search through cross-attention and multi-task training" [X Link](https://x.com/_reachsumit/status/1731560543545426097) 2023-12-04T06:26Z [---] followers, [---] engagements "Lookahead: An Inference Acceleration Framework for Large Language Model with Lossless Generation Accuracy Proposes Lookahead a multi-branch decoding framework using trie-based retrieval for accelerating inference of LLMs by overcoming I/O constraints" [X Link](https://x.com/_reachsumit/status/1738089567071416574) 2023-12-22T06:50Z [---] followers, [---] engagements "GUITAR: Gradient Pruning toward Fast Neural Ranking Baidu Research proposes a fast graph search method for neural ranking that uses gradient-based candidate pruning and an adaptive heuristic to minimize expensive neural network computations" [X Link](https://x.com/_reachsumit/status/1740597184280891642) 2023-12-29T04:54Z [---] followers, [---] engagements "Searching fast and slow through product catalogs Microsoft presents a fast and accurate SKU search system for CRMs combining Trie-based suggestions TF-IDF retrieval and language model embeddings outperforming existing systems" [X Link](https://x.com/_reachsumit/status/1742013360441610517) 2024-01-02T02:42Z [---] followers, 16K engagements "A case study of Generative AI in MSX Sales Copilot: Improving seller productivity with a real-time question-answering system for content recommendation Microsoft presents a real-time QA system that matches seller queries to sales documentation" [X Link](https://x.com/_reachsumit/status/1745293560335999157) 2024-01-11T03:56Z [---] followers, [---] engagements "ChatQA: Building GPT-4 Level Conversational QA Models Nvidia introduces a family of conversational QA models matching GPT-4 performance using efficient training methods like two-stage instruction tuning and improved context retrieval" [X Link](https://x.com/anyuser/status/1748168269541101901) 2024-01-19T02:19Z [--] followers, [---] engagements "In-context Learning with Retrieved Demonstrations for Language Models: A Survey Google Research offers a comprehensive analysis of retrieval-based ICL highlighting key innovations and future paths to enhance demonstration relevance and diversity" [X Link](https://x.com/_reachsumit/status/1749640183178514938) 2024-01-23T03:48Z [---] followers, [----] engagements "Enhancing Recommendation Diversity by Re-ranking with Large Language Models Shows that prompting LLMs to rerank recommendations for diversity shows promise but currently underperforms traditional diversification techniques" [X Link](https://x.com/_reachsumit/status/1749661743771685062) 2024-01-23T05:14Z [---] followers, [---] engagements "Multimodal Query Suggestion with Multi-Agent Reinforcement Learning from Human Feedback Huawei presents a new task to generate query suggestions from user images rather than just text to better capture intent and diversity" [X Link](https://x.com/_reachsumit/status/1755472891544740242) 2024-02-08T06:05Z [---] followers, [---] engagements "RA-Rec: An Efficient ID Representation Alignment Framework for LLM-based Recommendation Huawei presents a framework that aligns ID embeddings with LLMs for recommendation systems to leverage the strengths of both approaches" [X Link](https://x.com/_reachsumit/status/1755476567273451984) 2024-02-08T06:20Z [---] followers, [---] engagements "CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models Presents a LLM-powered plugin for Slack facilitating collaborative search within conversations by understanding context and providing summaries" [X Link](https://x.com/_reachsumit/status/1756864760082280940) 2024-02-12T02:16Z [---] followers, [----] engagements "Search Intention Network for Personalized Query Auto-Completion in E-Commerce Alibaba proposes a framework for query auto-completion that models a user's current intention from the prefix and historical intentions from behavior sequences" [X Link](https://x.com/_reachsumit/status/1765278458534670837) 2024-03-06T07:29Z [----] followers, [---] engagements "Aligning Large Language Models for Controllable Recommendations Microsoft proposes a two-stage supervised and RL approach to enhance LLMs' ability to follow instructions and reduce errors in recommender systems" [X Link](https://x.com/_reachsumit/status/1767052280434684251) 2024-03-11T04:57Z [----] followers, [---] engagements "MetaSplit: Meta-Split Network for Limited-Stock Product Recommendation Alibaba proposes addressing the challenges of recommending limited-stock products in C2C e-commerce platforms by employing differentiated modeling approaches based on stock levels" [X Link](https://x.com/_reachsumit/status/1767396053140316379) 2024-03-12T03:43Z [----] followers, [---] engagements "Detecting Hallucination and Coverage Errors in Retrieval Augmented Generation for Controversial Topics Google proposes using RAG with LLM-based classifiers trained on synthetic errors to handle controversial topics" [X Link](https://x.com/_reachsumit/status/1768458742394392602) 2024-03-15T02:06Z [----] followers, [---] engagements "SelfIE: Self-Interpretation of Large Language Model Embeddings Introduces a framework that enables LLMs to interpret their own hidden embeddings in natural language revealing their internal reasoning processes. https://selfie.cs.columbia.edu/ https://arxiv.org/abs/2403.10949 https://selfie.cs.columbia.edu/ https://arxiv.org/abs/2403.10949" [X Link](https://x.com/_reachsumit/status/1769933438814224679) 2024-03-19T03:46Z [----] followers, [---] engagements "AlphaFin: Benchmarking Financial Analysis with Retrieval-Augmented Stock-Chain Framework Alibaba proposes a framework that integrates LLMs with a real-time financial database to perform stock trend prediction and financial QA" [X Link](https://x.com/_reachsumit/status/1770304480342626647) 2024-03-20T04:20Z [----] followers, [---] engagements "CoLLEGe: Concept Embedding Generation for Large Language Models Presents a meta-learning framework that enables LLMs to quickly acquire new concepts from few examples by generating embeddings" [X Link](https://x.com/_reachsumit/status/1772079763500437785) 2024-03-25T01:55Z [----] followers, [----] engagements "MA4DIV: Multi-Agent Reinforcement Learning for Search Result Diversification Models search result diversification as a cooperative task among document agents directly optimizing diversity metrics while improving efficiency over existing methods. https://arxiv.org/abs/2403.17421 https://arxiv.org/abs/2403.17421" [X Link](https://x.com/_reachsumit/status/1772874886354268190) 2024-03-27T06:34Z [----] followers, [---] engagements "Make Large Language Model a Better Ranker Aligns large language models with listwise ranking objectives for recommender systems using soft lambda loss and permutation-sensitive learning to address computational cost and position bias challenges. https://arxiv.org/abs/2403.19181 https://arxiv.org/abs/2403.19181" [X Link](https://x.com/_reachsumit/status/1773567359854194999) 2024-03-29T04:26Z [----] followers, [---] engagements "FT2Ra: A Fine-Tuning-Inspired Approach to Retrieval-Augmented Code Completion Introduces a retrieval-based approach that mimics fine-tuning by iteratively augmenting a code pre-trained model's predictions using retrieved delta logits. https://arxiv.org/abs/2404.01554 https://arxiv.org/abs/2404.01554" [X Link](https://x.com/_reachsumit/status/1775411275826372880) 2024-04-03T06:33Z [----] followers, [---] engagements "Event-enhanced Retrieval in Real-time Search Tencent tackles the "semantic drift" problem in real-time search by incorporating a generative decoder to extract event-focused representations. https://github.com/open-event-hub/Event-enhanced_Retrieval https://arxiv.org/abs/2404.05989 https://github.com/open-event-hub/Event-enhanced_Retrieval https://arxiv.org/abs/2404.05989" [X Link](https://x.com/_reachsumit/status/1778292382892777557) 2024-04-11T05:21Z [----] followers, [---] engagements "Graph Chain-of-Thought: Augmenting Large Language Models by Reasoning on Graphs Introduces a benchmark for evaluating how LLMs can leverage KGs and proposes a framework that enables iterative reasoning on graphs. https://arxiv.org/abs/2404.07103 https://github.com/PeterGriffinJin/Graph-CoT https://arxiv.org/abs/2404.07103 https://github.com/PeterGriffinJin/Graph-CoT" [X Link](https://x.com/_reachsumit/status/1778296018200862727) 2024-04-11T05:36Z [----] followers, [---] engagements "Reducing hallucination in structured outputs via Retrieval-Augmented Generation ServiceNow proposes to reduce hallucinations and improve generalization in a system that converts natural language requirements into structured workflow representations. https://arxiv.org/abs/2404.08189 https://arxiv.org/abs/2404.08189" [X Link](https://x.com/_reachsumit/status/1779709247775031721) 2024-04-15T03:12Z [----] followers, [----] engagements "Navigating the Evaluation Funnel to Optimize Iteration Speed for Recommender Systems Spotify presents a framework for evaluating recommendation systems by decomposing the definition of success and combining offline and online evaluation methods. https://arxiv.org/abs/2404.08671 https://arxiv.org/abs/2404.08671" [X Link](https://x.com/_reachsumit/status/1780108545817821511) 2024-04-16T05:38Z [----] followers, [---] engagements "Retrieval Augmented Generation for Domain-specific Question Answering Adobe proposes a framework for developing an in-house QA system featuring a large QA database a retriever trained on domain data and user behavior and a finetuned language model. https://arxiv.org/abs/2404.14760 https://arxiv.org/abs/2404.14760" [X Link](https://x.com/_reachsumit/status/1782956221223600407) 2024-04-24T02:14Z [----] followers, [----] engagements "Hi-Gen: Generative Retrieval For Large-Scale Personalized E-commerce Search Alibaba encodes semantic information in document ID generation & exploits positional information during decoding to enable efficient large-scale personalized E-commerce search. https://arxiv.org/abs/2404.15675 https://arxiv.org/abs/2404.15675" [X Link](https://x.com/_reachsumit/status/1783388401691246899) 2024-04-25T06:51Z [----] followers, [----] engagements "OmniSearchSage: Multi-Task Multi-Entity Embeddings for Pinterest Search Pinterest jointly learns unified query pin and product embeddings using enriched entity representations. https://github.com/pinterest/atg-research/tree/main/omnisearchsage https://arxiv.org/abs/2404.16260 https://github.com/pinterest/atg-research/tree/main/omnisearchsage https://arxiv.org/abs/2404.16260" [X Link](https://x.com/_reachsumit/status/1783669365776466400) 2024-04-26T01:28Z [----] followers, [----] engagements "InspectorRAGet: An Introspection Platform for RAG Evaluation Enables comprehensive evaluation of retrieval-augmented language models through aggregate and instance-level analysis mixed metrics and annotator assessment. https://github.com/IBM/InspectorRAGet https://arxiv.org/abs/2404.17347 https://github.com/IBM/InspectorRAGet https://arxiv.org/abs/2404.17347" [X Link](https://x.com/_reachsumit/status/1784786828111102327) 2024-04-29T03:28Z [----] followers, [----] engagements "Interest Clock: Time Perception in Real-Time Streaming Recommendation System Douyin/Bytedance encodes users' time-aware preferences into personalized clock embeddings and employs Gaussian smoothing to capture dynamic user preferences over time. https://arxiv.org/abs/2404.19357 https://arxiv.org/abs/2404.19357" [X Link](https://x.com/_reachsumit/status/1785483483974074616) 2024-05-01T01:36Z [----] followers, [---] engagements ""In-Context Learning" or: How I learned to stop worrying and love "Applied Information Retrieval" Draws an analogy between ICL for LLMs and IR proposes IR techniques to improve few-shot example selection and enhance downstream genAI tasks. https://arxiv.org/abs/2405.01116 https://arxiv.org/abs/2405.01116" [X Link](https://x.com/_reachsumit/status/1786221307669127426) 2024-05-03T02:28Z [----] followers, [----] engagements "Semi-Parametric Retrieval via Binary Token Index Utilizes efficient binary token indexing alongside embedding-based indexing achieving superior retrieval accuracy while reducing indexing time and storage requirements https://github.com/jzhoubu/VDR https://arxiv.org/abs/2405.01924 https://github.com/jzhoubu/VDR https://arxiv.org/abs/2405.01924" [X Link](https://x.com/_reachsumit/status/1787367068393988528) 2024-05-06T06:21Z [----] followers, [---] engagements "Lifelong Knowledge Editing for LLMs with Retrieval-Augmented Continuous Prompt Learning Alibaba enables lifelong model editing for LLMs by converting knowledge into continuous prompts and employing dynamic prompt retrieval. https://arxiv.org/abs/2405.03279 https://arxiv.org/abs/2405.03279" [X Link](https://x.com/_reachsumit/status/1787690383788699762) 2024-05-07T03:46Z [----] followers, [---] engagements "R4: Reinforced Retriever-Reorder-Responder for Retrieval-Augmented Large Language Models Alibaba improves retrieval-augmented LLMs by learning optimal document orderings and enhancing document representations via RL and adversarial techniques. https://arxiv.org/abs/2405.02659 https://arxiv.org/abs/2405.02659" [X Link](https://x.com/_reachsumit/status/1787697598524166390) 2024-05-07T04:14Z [----] followers, [---] engagements "FlashBack:Efficient Retrieval-Augmented Language Modeling for Long Context Inference Appends retrieved documents to improve inference efficiency while maintaining performance through tunable Marking Tokens and LoRA fine-tuning. https://arxiv.org/abs/2405.04065 https://arxiv.org/abs/2405.04065" [X Link](https://x.com/_reachsumit/status/1788068513183392088) 2024-05-08T04:48Z [----] followers, [---] engagements "Knowledge Adaptation from Large Language Model to Recommendation for Practical Industrial Application Kuaishou employs a frozen LLM for content embedding extraction and connects the open-world and collaborative knowledge domains through an SSL task. https://arxiv.org/abs/2405.03988 https://arxiv.org/abs/2405.03988" [X Link](https://x.com/_reachsumit/status/1788069903683207200) 2024-05-08T04:54Z [----] followers, [---] engagements "Positional encoding is not the same as context: A study on positional encoding for Sequential recommendation Huawei analyzes positional encodings in transformer-based sequential recommendation systems proposing new encodings. https://github.com/researcher1741/Position_encoding_SRS https://arxiv.org/abs/2405.10436 https://github.com/researcher1741/Position_encoding_SRS https://arxiv.org/abs/2405.10436" [X Link](https://x.com/_reachsumit/status/1792387642564461017) 2024-05-20T02:51Z [----] followers, [----] engagements "Empowering Small-Scale Knowledge Graphs: A Strategy of Leveraging General-Purpose Knowledge Graphs for Enriched Embeddings Proposes a framework to enrich small domain-specific knowledge graphs with large general-purpose KGs. https://github.com/graphml-lab-pwr/empowering-small-scale-kg https://arxiv.org/abs/2405.10745 https://github.com/graphml-lab-pwr/empowering-small-scale-kg https://arxiv.org/abs/2405.10745" [X Link](https://x.com/_reachsumit/status/1792389469527744964) 2024-05-20T02:58Z [----] followers, [----] engagements "From Sora What We Can See: A Survey of Text-to-Video Generation Reviews the text-to-video generation domain centered around OpenAI's Sora model identifying remaining challenges and proposing future research directions. https://github.com/soraw-ai/Awesome-Text-to-Video-Generation https://arxiv.org/abs/2405.10674 https://github.com/soraw-ai/Awesome-Text-to-Video-Generation https://arxiv.org/abs/2405.10674" [X Link](https://x.com/_reachsumit/status/1792390561510682780) 2024-05-20T03:03Z [----] followers, [----] engagements "Diversifying by Intent in Recommender Systems Google proposes a probabilistic intent-based diversification framework for recommender systems that incorporates higher-level user intents to optimize long-term user experience. https://arxiv.org/abs/2405.12327 https://arxiv.org/abs/2405.12327" [X Link](https://x.com/_reachsumit/status/1793105042813411529) 2024-05-22T02:22Z [----] followers, [---] engagements "NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models Nvidia introduces an embedding model with architectural improvements like latent attention pooling and a 2-stage contrastive instruction-tuning. https://huggingface.co/nvidia/NV-Embed-v1 https://arxiv.org/abs/2405.17428 https://huggingface.co/nvidia/NV-Embed-v1 https://arxiv.org/abs/2405.17428" [X Link](https://x.com/_reachsumit/status/1795303361455202558) 2024-05-28T03:57Z [----] followers, [----] engagements "Generative Query Reformulation Using Ensemble Prompting Document Fusion and Relevance Feedback Proposes ensemble-based prompting techniques that leverage paraphrastic instructions to generate diverse query reformulations. https://arxiv.org/abs/2405.17658 https://arxiv.org/abs/2405.17658" [X Link](https://x.com/_reachsumit/status/1795676737726194101) 2024-05-29T04:41Z [----] followers, [---] engagements "RAG Does Not Work for Enterprises Explores the challenges and requirements for implementing RAG in enterprises proposing potential solutions like semantic search and hybrid queries and an evaluation framework to validate enterprise-grade RAG solutions https://arxiv.org/abs/2406.04369 https://arxiv.org/abs/2406.04369" [X Link](https://x.com/_reachsumit/status/1800007617420623956) 2024-06-10T03:30Z [----] followers, 107.1K engagements "Async Learned User Embeddings for Ads Delivery Optimization Meta learns high-fidelity user embeddings from multimodal user activities using a Transformer-like model and graph learning enabling effective ad retrieval. https://arxiv.org/abs/2406.05898 https://arxiv.org/abs/2406.05898" [X Link](https://x.com/_reachsumit/status/1800381901485883792) 2024-06-11T04:17Z [----] followers, [---] engagements "Optimizing Novelty of Top-k Recommendations using Large Language Models and Reinforcement Learning Microsoft optimizes novelty in large-scale recommendation systems using LLMs for feedback and a reformulated state-action space to improve efficiency. https://arxiv.org/abs/2406.14169 https://arxiv.org/abs/2406.14169" [X Link](https://x.com/_reachsumit/status/1804028096926076940) 2024-06-21T05:46Z [----] followers, [---] engagements "Do Not Wait: Learning Re-Ranking Model Without User Feedback At Serving Time in E-Commerce Generates transient request-specific model modifications using a surrogate model enabling immediate adaptation without relying on delayed user feedback. https://arxiv.org/abs/2406.14004 https://arxiv.org/abs/2406.14004" [X Link](https://x.com/_reachsumit/status/1804034631723028492) 2024-06-21T06:12Z [----] followers, [---] engagements "This post concludes the two-part series on the evolution of the MTL-based video recommender system. It describes disentanglement strategies to enable effective and efficient learning of inter-task relationships. https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/ https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/" [X Link](https://x.com/_reachsumit/status/1804568654576865785) 2024-06-22T17:34Z [----] followers, [---] engagements "Additionally the post emphasizes real-world solutions from Kuaishou Tencent YouTube Facebook and Amazon Prime Video to address different biases and shares tips from various published research works. https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/ https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/" [X Link](https://x.com/_reachsumit/status/1804568656225227255) 2024-06-22T17:34Z [----] followers, [---] engagements "FIRST: Faster Improved Listwise Reranking with Single Token Decoding Proposes an LLM reranking method that uses only the first token's logits to rank passages improving efficiency by 50% while maintaining performance. https://github.com/gangiswag/llm-reranker https://arxiv.org/abs/2406.15657 https://github.com/gangiswag/llm-reranker https://arxiv.org/abs/2406.15657" [X Link](https://x.com/_reachsumit/status/1805452468115128444) 2024-06-25T04:06Z [----] followers, [----] engagements "T-FREE: Tokenizer-Free Generative LLMs via Sparse Representations for Memory-Efficient Embeddings Presents a tokenizer-free approach for LLMs that directly embeds words using character triplets showing competitive performance https://github.com/Aleph-Alpha/trigrams https://arxiv.org/abs/2406.19223 https://github.com/Aleph-Alpha/trigrams https://arxiv.org/abs/2406.19223" [X Link](https://x.com/_reachsumit/status/1806533326121177319) 2024-06-28T03:41Z [----] followers, [---] engagements "Summary of a Haystack: A Challenge to Long-Context LLMs and RAG Systems Salesforce presents a new benchmark for evaluating long-context language models and RAG systems requiring precise summarization and citation of insights. https://github.com/salesforce/summary-of-a-haystack https://arxiv.org/abs/2407.01370 https://github.com/salesforce/summary-of-a-haystack https://arxiv.org/abs/2407.01370" [X Link](https://x.com/_reachsumit/status/1808168716703879335) 2024-07-02T15:59Z [----] followers, [----] engagements "RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs NVIDIA introduces an instruction fine-tuning framework that enables a single LLM to perform both context ranking and answer generation in RAG. https://arxiv.org/abs/2407.02485 https://arxiv.org/abs/2407.02485" [X Link](https://x.com/_reachsumit/status/1808522379787964556) 2024-07-03T15:25Z [----] followers, 20.2K engagements "MemoCRS: Memory-enhanced Sequential Conversational Recommender Systems with Large Language Models Huawei uses memory-enhanced LLMs to capture user preference continuity across dialogue sessions. https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/memocrs https://arxiv.org/abs/2407.04960 https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/memocrs https://arxiv.org/abs/2407.04960" [X Link](https://x.com/_reachsumit/status/1810543204611338608) 2024-07-09T05:15Z [----] followers, [---] engagements "Merge Ensemble and Cooperate A Survey on Collaborative Strategies in the Era of Large Language Models Examines collaboration strategies for LLMs categorizing them into Merging Ensemble and Cooperation approaches and provides reviews of each. https://arxiv.org/abs/2407.06089 https://arxiv.org/abs/2407.06089" [X Link](https://x.com/_reachsumit/status/1810545303390654509) 2024-07-09T05:23Z [----] followers, [---] engagements "RAG vs. Long Context: Examining Frontier Large Language Models for Environmental Review Document Comprehension Introduces a benchmark for evaluating LLMs' performance on NEPA documents finding that RAG-powered models outperform long-context LLMs. https://arxiv.org/abs/2407.07321 https://arxiv.org/abs/2407.07321" [X Link](https://x.com/_reachsumit/status/1811224451029336525) 2024-07-11T02:22Z [----] followers, [----] engagements "Deep Bag-of-Words Model: An Efficient and Interpretable Relevance Architecture for Chinese E-Commerce Alibaba combines deep learning with sparse Bag-of-Words representation improving performance interpretability and efficiency over dense models. https://arxiv.org/abs/2407.09395 https://arxiv.org/abs/2407.09395" [X Link](https://x.com/_reachsumit/status/1812673004310233405) 2024-07-15T02:18Z [----] followers, [----] engagements "ChatQA 2: Bridging the Gap to Proprietary LLMs in Long Context and RAG Capabilities NVIDIA extends Llama3-70B to rival GPT-4-Turbo in long-context understanding and RAG achieving comparable performance on various tasks. https://arxiv.org/abs/2407.14482 https://arxiv.org/abs/2407.14482" [X Link](https://x.com/_reachsumit/status/1815212629948068161) 2024-07-22T02:29Z [----] followers, [----] engagements "NV-Retriever: Improving text embedding models with effective hard-negative mining NVIDIA presents positive-aware hard-negative mining for text embedding models and a model that topped the MTEB Retrieval benchmark in July'24. https://huggingface.co/nvidia/NV-Retriever-v1 https://arxiv.org/abs/2407.15831 https://huggingface.co/nvidia/NV-Retriever-v1 https://arxiv.org/abs/2407.15831" [X Link](https://x.com/_reachsumit/status/1815622664071114765) 2024-07-23T05:39Z [----] followers, [----] engagements "Efficient Retrieval with Learned Similarities Introduces Mixture-of-Logits (MoL) as a universal approximator for learned similarity functions in retrieval tasks proposing efficient techniques for approximate top-K retrieval. https://arxiv.org/abs/2407.15462 https://arxiv.org/abs/2407.15462" [X Link](https://x.com/_reachsumit/status/1815624111181472113) 2024-07-23T05:44Z [----] followers, [----] engagements "TWIN V2: Scaling Ultra-Long User Behavior Sequence Modeling for Enhanced CTR Prediction at Kuaishou Kuaishou introduces a method to extend user behavior modeling to lifecycle scale (up to [---] interactions) for CTR prediction in recommendation systems. https://arxiv.org/abs/2407.16357 https://arxiv.org/abs/2407.16357" [X Link](https://x.com/_reachsumit/status/1815932577548755162) 2024-07-24T02:10Z [----] followers, [---] engagements "RazorAttention: Efficient KV Cache Compression Through Retrieval Heads Huawei proposes a KV cache compression technique compatible with FlashAttention for LLMs that reduces cache size by over 70% without significant performance loss. https://arxiv.org/abs/2407.15891 https://arxiv.org/abs/2407.15891" [X Link](https://x.com/_reachsumit/status/1815933735898079365) 2024-07-24T02:15Z [----] followers, [---] engagements "Improving Retrieval Augmented Language Model with Self-Reasoning Baidu enhances RALMs by using LLM-generated reasoning trajectories improving reliability and traceability rivaling GPT-4's performance on QA tasks using minimal training data. https://arxiv.org/abs/2407.19813 https://arxiv.org/abs/2407.19813" [X Link](https://x.com/_reachsumit/status/1818131488988348611) 2024-07-30T03:48Z [----] followers, [----] engagements "Enhancing Taobao Display Advertising with Multimodal Representations: Challenges Approaches and Insights Alibaba presents a two-phase framework for integrating multimodal data into large-scale recommendation systems. https://arxiv.org/abs/2407.19467 https://arxiv.org/abs/2407.19467" [X Link](https://x.com/_reachsumit/status/1818143296129991147) 2024-07-30T04:35Z [----] followers, [---] engagements "Breaking the Hourglass Phenomenon of Residual Quantization: Enhancing the Upper Bound of Generative Retrieval Identifies the "Hourglass" phenomenon in generative retrieval proposing solutions to mitigate token concentration in intermediate layers. https://arxiv.org/abs/2407.21488 https://arxiv.org/abs/2407.21488" [X Link](https://x.com/_reachsumit/status/1818834941834338569) 2024-08-01T02:23Z [----] followers, [---] engagements "MLLM Is a Strong Reranker: Advancing Multimodal Retrieval-augmented Generation via Knowledge-enhanced Reranking and Noise-injected Training Presents a framework that addresses the multi-granularity noisy correspondence problem https://github.com/IDEA-FinAI/RagLLaVA https://arxiv.org/abs/2407.21439 https://github.com/IDEA-FinAI/RagLLaVA https://arxiv.org/abs/2407.21439" [X Link](https://x.com/_reachsumit/status/1818835939583836513) 2024-08-01T02:27Z [----] followers, [----] engagements "Simple but Efficient: A Multi-Scenario Nearline Retrieval Framework for Recommendation on Taobao Alibaba introduces a framework that incorporates ranking results from various scenarios into the matching stage demonstrating a 5% increase in transactions https://arxiv.org/abs/2408.00247 https://arxiv.org/abs/2408.00247" [X Link](https://x.com/_reachsumit/status/1819200959597498490) 2024-08-02T02:38Z [----] followers, [---] engagements "RECE: Reduced Cross-Entropy Loss for Large-Catalogue Sequential Recommenders Presents a loss function that approximates CE loss for sequential recommenders significantly reducing GPU memory usage while maintaining performance https://github.com/dalibra/RECE https://arxiv.org/abs/2408.02354 https://github.com/dalibra/RECE https://arxiv.org/abs/2408.02354" [X Link](https://x.com/_reachsumit/status/1820683184394400140) 2024-08-06T04:47Z [----] followers, [---] engagements "A Real-Time Adaptive Multi-Stream GPU System for Online Approximate Nearest Neighborhood Search Xiaohongshu presents a GPU-based system for real-time Approximate Nearest Neighbor Search featuring dynamic vector insertion and multi-stream execution. https://arxiv.org/abs/2408.02937 https://arxiv.org/abs/2408.02937" [X Link](https://x.com/_reachsumit/status/1821004375302758820) 2024-08-07T02:04Z [----] followers, [----] engagements "Relevance meets Diversity: A User-Centric Framework for Knowledge Exploration through Recommendations Introduces a user-centric recommender system that maximizes knowledge acquisition by balancing relevance and diversity. https://github.com/EricaCoppolillo/EXPLORE https://arxiv.org/abs/2408.03772 https://github.com/EricaCoppolillo/EXPLORE https://arxiv.org/abs/2408.03772" [X Link](https://x.com/_reachsumit/status/1821375927647195425) 2024-08-08T02:40Z [----] followers, [----] engagements "Advancing Re-Ranking with Multimodal Fusion and Target-Oriented Auxiliary Tasks in E-Commerce Search JD.com presents a new e-commerce search re-ranking model that integrates multimodal information using attention-based fusion and auxiliary tasks. https://arxiv.org/abs/2408.05751 https://arxiv.org/abs/2408.05751" [X Link](https://x.com/_reachsumit/status/1823212039264665628) 2024-08-13T04:16Z [----] followers, [---] engagements "HoME: Hierarchy of Multi-Gate Experts for Multi-Task Learning at Kuaishou Kuaishou presents a multi-task learning framework for short-video recommendations that addresses expert collapse degradation and underfitting in MoE systems. https://arxiv.org/abs/2408.05430 https://arxiv.org/abs/2408.05430" [X Link](https://x.com/_reachsumit/status/1823214370186539253) 2024-08-13T04:25Z [----] followers, [---] engagements "Mamba Retriever: Utilizing Mamba for Effective and Efficient Dense Retrieval Presents a dense retrieval model based on the Mamba architecture offering superior efficiency for long-text retrieval due to its linear time scaling https://github.com/41924076/MambaRetriever https://arxiv.org/abs/2408.08066 https://github.com/41924076/MambaRetriever https://arxiv.org/abs/2408.08066" [X Link](https://x.com/_reachsumit/status/1824276029646704666) 2024-08-16T02:44Z [----] followers, [----] engagements "ColBERT's MASK-based Query Augmentation: Effects of Quadrupling the Query Input Length While query effectiveness is optimized at a specific MASK token count the model remains robust even with significantly more tokens than it was trained on. https://arxiv.org/abs/2408.13672 https://arxiv.org/abs/2408.13672" [X Link](https://x.com/_reachsumit/status/1828292651092742234) 2024-08-27T04:45Z [----] followers, [----] engagements "LRP4RAG: Detecting Hallucinations in Retrieval-Augmented Generation via Layer-wise Relevance Propagation Analyzes relevance between input and output and uses Layer-wise Relevance Propagation to detect hallucinations in RAG systems. https://arxiv.org/abs/2408.15533 https://arxiv.org/abs/2408.15533" [X Link](https://x.com/_reachsumit/status/1828986981524869121) 2024-08-29T02:44Z [----] followers, [---] engagements "CRAFT Your Dataset: Task-Specific Synthetic Dataset Generation Through Corpus Retrieval and Augmentation Presents a method for generating task-specific synthetic datasets using user-provided few-shot examples. https://github.com/ziegler-ingo/CRAFT https://arxiv.org/abs/2409.02098 https://github.com/ziegler-ingo/CRAFT https://arxiv.org/abs/2409.02098" [X Link](https://x.com/_reachsumit/status/1831199648574239068) 2024-09-04T05:16Z [----] followers, [----] engagements "Genetic Approach to Mitigate Hallucination in Generative IR Introduces a method using a genetic algorithm with a balanced fitness function to reduce hallucinations in generative language models. https://github.com/Georgetown-IR-Lab/GAuGE https://arxiv.org/abs/2409.00085 https://github.com/Georgetown-IR-Lab/GAuGE https://arxiv.org/abs/2409.00085" [X Link](https://x.com/_reachsumit/status/1831207969570382129) 2024-09-04T05:49Z [----] followers, [---] engagements "Building a Scalable Effective and Steerable Search and Ranking Platform Zalando introduces a scalable real-time e-commerce ranking platform using transformer-based models to personalize product recommendations across various use cases. https://arxiv.org/abs/2409.02856 https://arxiv.org/abs/2409.02856" [X Link](https://x.com/_reachsumit/status/1831529189339762825) 2024-09-05T03:06Z [----] followers, [----] engagements "RouterRetriever: Exploring the Benefits of Routing over Multiple Expert Embedding Models Introduces a flexible multi-expert approach to information retrieval that routes queries to domain-specific experts. https://github.com/amy-hyunji/RouterRetriever https://arxiv.org/abs/2409.02685 https://github.com/amy-hyunji/RouterRetriever https://arxiv.org/abs/2409.02685" [X Link](https://x.com/_reachsumit/status/1831531095311184286) 2024-09-05T03:13Z [----] followers, [----] engagements "Diversify-verify-adapt: Efficient and Robust Retrieval-Augmented Ambiguous Question Answering Proposes a framework for handling ambiguous questions in RAG system using retrieval diversification and adaptive generation to improve accuracy and efficiency https://arxiv.org/abs/2409.02361 https://arxiv.org/abs/2409.02361" [X Link](https://x.com/_reachsumit/status/1831532164879642703) 2024-09-05T03:17Z [----] followers, [---] engagements "MOBIUS: Towards the Next Generation of Query-Ad Matching in Baidu's Sponsored Search Baidu introduces a unified query-ad matching system that considers both relevance and commercial performance utilizing active learning and efficient ANN techniques. https://arxiv.org/abs/2409.03449 https://arxiv.org/abs/2409.03449" [X Link](https://x.com/_reachsumit/status/1831903214800916811) 2024-09-06T03:52Z [----] followers, [---] engagements "Large Language Model-Based Agents for Software Engineering: A Survey Analyzes [---] papers on LLM-based agents in Software Engineering examining their applications across various development tasks and agent designs. https://github.com/FudanSELab/Agent4SE-Paper-List https://arxiv.org/abs/2409.02977 https://github.com/FudanSELab/Agent4SE-Paper-List https://arxiv.org/abs/2409.02977" [X Link](https://x.com/_reachsumit/status/1831903907637031142) 2024-09-06T03:55Z [----] followers, [----] engagements "OneGen: Efficient One-Pass Unified Generation and Retrieval for LLMs Presents a framework that enables LLMs to perform both generation and retrieval in a single forward pass. https://github.com/zjunlp/OneGen https://arxiv.org/abs/2409.05152 https://github.com/zjunlp/OneGen https://arxiv.org/abs/2409.05152" [X Link](https://x.com/_reachsumit/status/1833365133071991083) 2024-09-10T04:41Z [----] followers, [----] engagements "Late Chunking: Contextual Chunk Embeddings Using Long-Context Embedding Models Jina AI presents a technique that improves text embeddings for retrieval tasks by encoding entire documents before splitting them. https://github.com/jina-ai/late-chunking https://arxiv.org/abs/2409.04701 https://github.com/jina-ai/late-chunking https://arxiv.org/abs/2409.04701" [X Link](https://x.com/_reachsumit/status/1833367390899593366) 2024-09-10T04:50Z [----] followers, [----] engagements "LexBoost: Improving Lexical Document Retrieval with Nearest Neighbors Uses pre-computed document neighborhoods to boost relevance scores and enhance lexical search. https://github.com/Georgetown-IR-Lab/LexBoost https://arxiv.org/abs/2409.05882 https://github.com/Georgetown-IR-Lab/LexBoost https://arxiv.org/abs/2409.05882" [X Link](https://x.com/_reachsumit/status/1833699513632256305) 2024-09-11T02:50Z [----] followers, [----] engagements "RePlay: a Recommendation Framework for Experimentation and Production Use Presents an open-source toolkit for building recommender systems that supports multiple data processing backends and hardware architectures. https://github.com/sb-ai-lab/RePlay https://arxiv.org/abs/2409.07272 https://github.com/sb-ai-lab/RePlay https://arxiv.org/abs/2409.07272" [X Link](https://x.com/_reachsumit/status/1834059256980742638) 2024-09-12T02:39Z [----] followers, [----] engagements "Negative Sampling in Recommendation: A Survey and Future Directions Comprehensively reviews negative sampling strategies in recommender systems categorizing existing methods discussing challenges and future directions. https://github.com/hulkima/NS4RS https://arxiv.org/abs/2409.07237 https://github.com/hulkima/NS4RS https://arxiv.org/abs/2409.07237" [X Link](https://x.com/_reachsumit/status/1834061782609965534) 2024-09-12T02:49Z [----] followers, [---] engagements "Enhancing Q&A Text Retrieval with Ranking Models: Benchmarking fine-tuning and deploying Rerankers for RAG NVIDIA benchmarks ranking models for text retrieval in QA tasks introducing a new sota model NV-RerankQA-Mistral-4B-v3 https://build.nvidia.com/explore/retrieval#nv-rerankqa-mistral-4b-v3 https://arxiv.org/abs/2409.07691 https://build.nvidia.com/explore/retrieval#nv-rerankqa-mistral-4b-v3 https://arxiv.org/abs/2409.07691" [X Link](https://x.com/_reachsumit/status/1834437783085068795) 2024-09-13T03:43Z [----] followers, [----] engagements "Agents in Software Engineering: Survey Landscape and Vision Presents a comprehensive survey on LLM-based agents in software engineering introducing a framework with perception memory and action modules. https://github.com/DeepSoftwareAnalytics/Awesome-Agent4SE https://arxiv.org/abs/2409.09030 https://github.com/DeepSoftwareAnalytics/Awesome-Agent4SE https://arxiv.org/abs/2409.09030" [X Link](https://x.com/_reachsumit/status/1835535376997654587) 2024-09-16T04:25Z [----] followers, [----] engagements "RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval Accelerates attention computation in LLMs by using a vector search-based approach to retrieve key-value pairs from CPU memory. https://github.com/mit-han-lab/quest https://arxiv.org/abs/2409.10516 https://github.com/mit-han-lab/quest https://arxiv.org/abs/2409.10516" [X Link](https://x.com/_reachsumit/status/1835892435765055494) 2024-09-17T04:04Z [----] followers, [----] engagements "beeFormer: Bridging the Gap Between Semantic and Interaction Similarity in Recommender Systems Introduces a framework for training sentence Transformer models on interaction data with text-side information. https://github.com/recombee/beeformer https://arxiv.org/abs/2409.10309 https://github.com/recombee/beeformer https://arxiv.org/abs/2409.10309" [X Link](https://x.com/_reachsumit/status/1835894438742401132) 2024-09-17T04:11Z [----] followers, [---] engagements "Trustworthiness in Retrieval-Augmented Generation Systems: A Survey Proposes a unified framework to evaluate the trustworthiness of RAG systems across six dimensions offering benchmarks and insights. https://github.com/smallporridge/TrustworthyRAG https://arxiv.org/abs/2409.10102 https://github.com/smallporridge/TrustworthyRAG https://arxiv.org/abs/2409.10102" [X Link](https://x.com/_reachsumit/status/1835896619335897544) 2024-09-17T04:20Z [----] followers, [----] engagements "HyPA-RAG: A Hybrid Parameter Adaptive Retrieval-Augmented Generation System for AI Legal and Policy Applications Presents a hybrid RAG system with adaptive parameter tuning and combined retrieval methods improving accuracy and fidelity. https://arxiv.org/abs/2409.09046 https://arxiv.org/abs/2409.09046" [X Link](https://x.com/_reachsumit/status/1835902608155840915) 2024-09-17T04:44Z [----] followers, [---] engagements "Promptriever: Instruction-Trained Retrievers Can Be Prompted Like Language Models Proposes a retrieval model that follows natural language instructions enabling more flexible and user-friendly search experiences. https://github.com/orionw/promptriever https://arxiv.org/abs/2409.11136 https://github.com/orionw/promptriever https://arxiv.org/abs/2409.11136" [X Link](https://x.com/_reachsumit/status/1836267386812665874) 2024-09-18T04:53Z [----] followers, [----] engagements "Measuring and Enhancing Trustworthiness of LLMs in RAG through Grounded Attributions and Learning to Refuse Introduces a metric for evaluating LLM trustworthiness in RAG systems and a framework to improve grounded responses. https://github.com/declare-lab/trust-align https://arxiv.org/abs/2409.11242 https://github.com/declare-lab/trust-align https://arxiv.org/abs/2409.11242" [X Link](https://x.com/_reachsumit/status/1836269738181468498) 2024-09-18T05:03Z [----] followers, [----] engagements "Retrieve Annotate Evaluate Repeat: Leveraging Multimodal LLMs for Large-Scale Product Retrieval Evaluation Zalando presents a framework using Multimodal LLMs to efficiently evaluate large-scale e-commerce product retrieval systems. https://arxiv.org/abs/2409.11860 https://arxiv.org/abs/2409.11860" [X Link](https://x.com/_reachsumit/status/1836597647869907037) 2024-09-19T02:46Z [----] followers, [----] engagements "Fact Fetch and Reason: A Unified Evaluation of Retrieval-Augmented Generation Introduces an evaluation dataset designed to test RAG systems' capabilities in factual accuracy retrieval and reasoning. https://huggingface.co/datasets/google/frames-benchmark https://arxiv.org/abs/2409.12941 https://huggingface.co/datasets/google/frames-benchmark https://arxiv.org/abs/2409.12941" [X Link](https://x.com/_reachsumit/status/1836967053917364709) 2024-09-20T03:14Z [----] followers, [----] engagements "HLLM: Enhancing Sequential Recommendations via Hierarchical Large Language Models for Item and User Modeling ByteDance introduces a Hierarchical LLM architecture for sequential recommendation systems. https://github.com/bytedance/HLLM https://arxiv.org/abs/2409.12740 https://github.com/bytedance/HLLM https://arxiv.org/abs/2409.12740" [X Link](https://x.com/_reachsumit/status/1836968775683297312) 2024-09-20T03:21Z [----] followers, [----] engagements "Retrieval-Augmented Test Generation: How Far Are We Investigates the use of RAG for unit test generation comparing different knowledge sources like API documentation GitHub issues and StackOverflow Q&As. https://anonymous.4open.science/r/api_guided_testgen-FB88/README.md https://arxiv.org/abs/2409.12682 https://anonymous.4open.science/r/api_guided_testgen-FB88/README.md https://arxiv.org/abs/2409.12682" [X Link](https://x.com/_reachsumit/status/1836970732120584646) 2024-09-20T03:28Z [----] followers, [---] engagements "RAD-Bench: Evaluating Large Language Models Capabilities in Retrieval Augmented Dialogues Introduces a benchmark designed to evaluate LLMs in multi-turn dialogues using RAG focusing on retrieval synthesis and reasoning. https://github.com/mtkresearch/RAD-Bench https://arxiv.org/abs/2409.12558 https://github.com/mtkresearch/RAD-Bench https://arxiv.org/abs/2409.12558" [X Link](https://x.com/_reachsumit/status/1836971746986967191) 2024-09-20T03:32Z [----] followers, [---] engagements "Data Augmentation for Sequential Recommendation: A Survey Comprehensively reviews data augmentation methods for sequential recommendation systems. https://github.com/KingGugu/DA-CL-4Rec https://arxiv.org/abs/2409.13545 https://github.com/KingGugu/DA-CL-4Rec https://arxiv.org/abs/2409.13545" [X Link](https://x.com/_reachsumit/status/1838077670233612608) 2024-09-23T04:47Z [----] followers, [----] engagements "Contextual Compression in Retrieval-Augmented Generation for Large Language Models: A Survey Examines contextual compression techniques for LLMs focusing on their use in RAG systems and also presents a new taxonomy. https://github.com/SrGrace/Contextual-Compression https://arxiv.org/abs/2409.13385 https://github.com/SrGrace/Contextual-Compression https://arxiv.org/abs/2409.13385" [X Link](https://x.com/_reachsumit/status/1838078426655367435) 2024-09-23T04:50Z [----] followers, [----] engagements "Retrieval Augmented Generation (RAG) and Beyond: A Comprehensive Survey on How to Make your LLMs use External Data More Wisely Microsoft categorizes data-augmented LLM queries and proposes strategies to tackle challenges in specialized domains. https://arxiv.org/abs/2409.14924 https://arxiv.org/abs/2409.14924" [X Link](https://x.com/_reachsumit/status/1838436889487118838) 2024-09-24T04:34Z [----] followers, 14.2K engagements "Exploring Hint Generation Approaches in Open-Domain Question Answering Presents a context preparation method for QA systems that uses automatic hint generation instead of traditional retrieval or generation approaches. https://github.com/DataScienceUIBK/HintQA https://arxiv.org/abs/2409.16096 https://github.com/DataScienceUIBK/HintQA https://arxiv.org/abs/2409.16096" [X Link](https://x.com/_reachsumit/status/1838772595124281392) 2024-09-25T02:48Z [----] followers, [---] engagements "Making Text Embedders Few-Shot Learners Introduces a text embedding model that leverages LLMs' in-context learning capabilities to generate high-quality adaptable embeddings. https://github.com/FlagOpen/FlagEmbedding https://arxiv.org/abs/2409.15700 https://github.com/FlagOpen/FlagEmbedding https://arxiv.org/abs/2409.15700" [X Link](https://x.com/_reachsumit/status/1838774217682399430) 2024-09-25T02:55Z [----] followers, [---] engagements "Enhancing Performance and Scalability of Large-Scale Recommendation Systems with Jagged Flash Attention Meta introduces Jagged Interaction Kernels and Jagged Flash Attention optimizing recommender systems for variable-length categorical features. https://arxiv.org/abs/2409.15373 https://arxiv.org/abs/2409.15373" [X Link](https://x.com/_reachsumit/status/1838777972612043271) 2024-09-25T03:10Z [----] followers, [---] engagements "Block-Attention for Low-Latency RAG Significantly reduces inference latency in RAG by pre-computing and caching key-value states for independent blocks of input achieving comparable accuracy to traditional models. https://github.com/TemporaryLoRA/Block-Attention https://arxiv.org/abs/2409.15355 https://github.com/TemporaryLoRA/Block-Attention https://arxiv.org/abs/2409.15355" [X Link](https://x.com/_reachsumit/status/1838778703461126504) 2024-09-25T03:13Z [----] followers, [---] engagements "FusionANNS: An Efficient CPU/GPU Cooperative Processing Architecture for Billion-scale Approximate Nearest Neighbor Search Presents a high-performance ANNS system for billion-scale datasets that uses CPU/GPU collaboration and SSD storage. https://arxiv.org/abs/2409.16576c https://arxiv.org/abs/2409.16576c" [X Link](https://x.com/_reachsumit/status/1839134449323356646) 2024-09-26T02:46Z [----] followers, [----] engagements "Disentangling Questions from Query Generation for Task-Adaptive Retrieval Presents a query generation system that adapts to diverse search intents using meta-prompts and retriever feedback. https://github.com/lilys012/metaprompt-QG https://arxiv.org/abs/2409.16570 https://github.com/lilys012/metaprompt-QG https://arxiv.org/abs/2409.16570" [X Link](https://x.com/_reachsumit/status/1839135283905966177) 2024-09-26T02:49Z [----] followers, [---] engagements "Mixed-Precision Embeddings for Large-Scale Recommendation Models Presents a method for compressing embedding tables in recommender systems assigning varying precisions to feature groups based on importance. https://github.com/Leopold1423/mpe https://arxiv.org/abs/2409.20305 https://github.com/Leopold1423/mpe https://arxiv.org/abs/2409.20305" [X Link](https://x.com/_reachsumit/status/1841288516518744473) 2024-10-02T01:26Z [----] followers, [----] engagements "Crafting Personalized Agents through Retrieval-Augmented Generation on Editable Memory Graphs Huawei proposes creating personalized AI agents that leverage users' smartphone memories to enhance LLM capabilities. https://arxiv.org/abs/2409.19401 https://arxiv.org/abs/2409.19401" [X Link](https://x.com/_reachsumit/status/1841295191422173217) 2024-10-02T01:52Z [----] followers, [---] engagements "Winning Solution For Meta KDD Cup' [--] Describes the winning solutions for the KDD Cup [--] RAG challenge detailing a web retrieval framework with tuned LLMs and a regularized API approach achieving first place in all [--] tasks. https://gitlab.aicrowd.com/jiazunchen/kdd2024cup-crag-db3 https://arxiv.org/abs/2410.00005 https://gitlab.aicrowd.com/jiazunchen/kdd2024cup-crag-db3 https://arxiv.org/abs/2410.00005" [X Link](https://x.com/_reachsumit/status/1841307587008156106) 2024-10-02T02:41Z [----] followers, [---] engagements "Retro-li: Small-Scale Retrieval Augmented Generation Supporting Noisy Similarity Searches and Domain Shift Generalization Improves RAG systems using a small-scale database with enhanced semantic search and noise regularization https://github.com/IBM/Retrieval-Enhanced-Transformer-Little https://arxiv.org/abs/2410.00004 https://github.com/IBM/Retrieval-Enhanced-Transformer-Little https://arxiv.org/abs/2410.00004" [X Link](https://x.com/_reachsumit/status/1841308460287459665) 2024-10-02T02:45Z [----] followers, [---] engagements "Open-RAG: Enhanced Retrieval-Augmented Reasoning with Open-Source Large Language Models Enhances reasoning in RAG using open-source LLMs by transforming them into sparse mixture of experts models with adaptive retrieval. https://openragmoe.github.io/ https://arxiv.org/abs/2410.01782 https://openragmoe.github.io/ https://arxiv.org/abs/2410.01782" [X Link](https://x.com/_reachsumit/status/1841679172768502027) 2024-10-03T03:18Z [----] followers, [---] engagements "Elaborative Subtopic Query Reformulation for Broad and Indirect Queries in Travel Destination Recommendation Introduces an LLM-based query reformulation method that improves travel recommender systems. https://github.com/YifanLiu2/ROEGEN-RecSys-24-EQR https://arxiv.org/abs/2410.01598 https://github.com/YifanLiu2/ROEGEN-RecSys-24-EQR https://arxiv.org/abs/2410.01598" [X Link](https://x.com/_reachsumit/status/1841683051220500541) 2024-10-03T03:33Z [----] followers, [----] engagements "Contextual Document Embeddings Improves neural retrieval by incorporating document context through a new contrastive learning objective and a context-aware architecture. https://arxiv.org/abs/2410.02525 https://arxiv.org/abs/2410.02525" [X Link](https://x.com/_reachsumit/status/1842060241481171024) 2024-10-04T04:32Z [----] followers, [----] engagements "Dreaming User Multimodal Representation for Micro-Video Recommendation Kuaishou models user interests in a unified multimodal space leveraging historical interactions and addressing cold-start scenarios. https://arxiv.org/abs/2410.03538 https://arxiv.org/abs/2410.03538" [X Link](https://x.com/_reachsumit/status/1843124051939934424) 2024-10-07T02:59Z [----] followers, [---] engagements "Inductive Generative Recommendation via Retrieval-based Speculation Enables generative recommendation models to recommend new unseen items by combining an inductive drafter with a GR model verifier. https://github.com/Jamesding000/SpecGR https://arxiv.org/abs/2410.02939 https://github.com/Jamesding000/SpecGR https://arxiv.org/abs/2410.02939" [X Link](https://x.com/_reachsumit/status/1843129502211461514) 2024-10-07T03:21Z [----] followers, [----] engagements "Deciphering the Interplay of Parametric and Non-parametric Memory in Retrieval-augmented Language Models Examines how the Atlas RAG model balances parametric and non-parametric knowledge finding it favors retrieved context. https://github.com/m3hrdadfi/rag-memory-interplay https://arxiv.org/abs/2410.05162 https://github.com/m3hrdadfi/rag-memory-interplay https://arxiv.org/abs/2410.05162" [X Link](https://x.com/_reachsumit/status/1843528470938431756) 2024-10-08T05:46Z [----] followers, [---] engagements "TableRAG: Million-Token Table Understanding with Language Models Enables efficient large-scale table understanding for language models using smart retrieval techniques to overcome context length limitations while reducing token consumption. https://arxiv.org/abs/2410.04739 https://arxiv.org/abs/2410.04739" [X Link](https://x.com/_reachsumit/status/1843530517226062128) 2024-10-08T05:55Z [----] followers, [----] engagements "Enhancing Playback Performance in Video Recommender Systems with an On-Device Gating and Ranking Framework Kuaishou introduces an on-device Gating and Ranking Framework to address choppy video playback in recommender systems. https://arxiv.org/abs/2410.05863 https://arxiv.org/abs/2410.05863" [X Link](https://x.com/_reachsumit/status/1844271963247870136) 2024-10-10T07:01Z [----] followers, [---] engagements "Meta Learning to Rank for Sparsely Supervised Queries Presents a meta-learning to rank framework that effectively addresses sparsely supervised queries enabling quick adaptation to new queries through its ability to generate query-specific rankers. https://dl.acm.org/doi/10.1145/3698876 https://dl.acm.org/doi/10.1145/3698876" [X Link](https://x.com/_reachsumit/status/1844589112931881313) 2024-10-11T04:01Z [----] followers, [----] engagements "Retriever-and-Memory: Towards Adaptive Note-Enhanced Retrieval-Augmented Generation Uses iterative information collection and adaptive memory review to improve knowledge integration and answer quality. https://github.com/thunlp/Adaptive-Note https://arxiv.org/abs/2410.08821 https://github.com/thunlp/Adaptive-Note https://arxiv.org/abs/2410.08821" [X Link](https://x.com/_reachsumit/status/1845651557318320218) 2024-10-14T02:23Z [----] followers, [----] engagements "VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents Introduces a VLM-based RAG system that processes multi-modal documents as images achieving up to 39% performance improvement in retrieval tasks. https://github.com/openbmb/visrag https://arxiv.org/abs/2410.10594 https://github.com/openbmb/visrag https://arxiv.org/abs/2410.10594" [X Link](https://x.com/_reachsumit/status/1846047188432179487) 2024-10-15T04:35Z [----] followers, [---] engagements "Learning to Rank for Multiple Retrieval-Augmented Models through Iterative Utility Maximization Presents an iterative framework for optimizing a unified search engine serving multiple RAG agents. https://github.com/alirezasalemi7/uRAG https://arxiv.org/abs/2410.09942 https://github.com/alirezasalemi7/uRAG https://arxiv.org/abs/2410.09942" [X Link](https://x.com/_reachsumit/status/1846052250046910506) 2024-10-15T04:55Z [----] followers, [---] engagements "Agentic Information Retrieval Proposes a paradigm using LLM agents to expand and transform traditional information retrieval offering a unified flexible approach to complex information tasks. https://arxiv.org/abs/2410.09713 https://arxiv.org/abs/2410.09713" [X Link](https://x.com/_reachsumit/status/1846054685956100202) 2024-10-15T05:05Z [----] followers, [----] engagements "Synthetic Knowledge Ingestion: Towards Knowledge Refinement and Injection for Enhancing Large Language Models Intuit enhances LLMs' knowledge capabilities through fine-grained synthesis interleaved generation and assemble augmentation. https://arxiv.org/abs/2410.09629 https://arxiv.org/abs/2410.09629" [X Link](https://x.com/_reachsumit/status/1846055645864439945) 2024-10-15T05:08Z [----] followers, [---] engagements "Toward General Instruction-Following Alignment for Retrieval-Augmented Generation Introduces an automated pipeline for improving instruction-following in RAG systems alongside FollowRAG a new benchmark for evaluation. https://followrag.github.io/ https://arxiv.org/abs/2410.09584 https://followrag.github.io/ https://arxiv.org/abs/2410.09584" [X Link](https://x.com/_reachsumit/status/1846056508892795216) 2024-10-15T05:12Z [----] followers, [---] engagements "Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free Combines routing weights and hidden states from MoE LLMs to create superior embeddings without additional training. https://github.com/tianyi-lab/MoE-Embedding https://arxiv.org/abs/2410.10814 https://github.com/tianyi-lab/MoE-Embedding https://arxiv.org/abs/2410.10814" [X Link](https://x.com/_reachsumit/status/1846058130633117819) 2024-10-15T05:18Z [----] followers, 12.9K engagements "Sequential LLM Framework for Fashion Recommendation Presents a sequential fashion recommendation system using an LLM employing specialized prompts efficient fine-tuning and a mix-up-based retrieval technique. https://arxiv.org/abs/2410.11327 https://arxiv.org/abs/2410.11327" [X Link](https://x.com/_reachsumit/status/1846420032529879323) 2024-10-16T05:16Z [----] followers, [---] engagements "FRAG: Toward Federated Vector Database Management for Collaborative and Secure Retrieval-Augmented Generation Enables secure efficient collaboration by allowing mutually distrusting parties to perform encrypted ANN searche across distributed databases https://arxiv.org/abs/2410.13272 https://arxiv.org/abs/2410.13272" [X Link](https://x.com/_reachsumit/status/1847161431617318958) 2024-10-18T06:22Z [----] followers, [---] engagements "Starbucks: Improved Training for 2D Matryoshka Embeddings Improves 2D Matryoshka by using targeted layer-dimension pairs achieving effectiveness comparable to separately trained models while maintaining adaptability. https://github.com/ielab/Starbucks https://arxiv.org/abs/2410.13230 https://github.com/ielab/Starbucks https://arxiv.org/abs/2410.13230" [X Link](https://x.com/_reachsumit/status/1847163736605515975) 2024-10-18T06:32Z [----] followers, [----] engagements "Optimizing and Evaluating Enterprise Retrieval-Augmented Generation (RAG): A Content Design Perspective IBM details practical experiences in building enterprise-scale RAG solutions for software documentation. https://github.com/spackows/ICAAI-2024_RAG-CD https://arxiv.org/abs/2410.12812 https://github.com/spackows/ICAAI-2024_RAG-CD https://arxiv.org/abs/2410.12812" [X Link](https://x.com/_reachsumit/status/1847167220834881646) 2024-10-18T06:45Z [----] followers, [---] engagements "Simplify to the Limit Embedding-less Graph Collaborative Filtering for Recommender Systems Improves on GCN and GCL approaches by focusing on user-item similarity eliminating embeddings and using contrastive objectives. https://github.com/BlueGhostYi/ID-GRec https://dl.acm.org/doi/10.1145/3701230 https://github.com/BlueGhostYi/ID-GRec https://dl.acm.org/doi/10.1145/3701230" [X Link](https://x.com/_reachsumit/status/1848194675829535150) 2024-10-21T02:48Z [----] followers, [----] engagements "Centrality-aware Product Retrieval and Ranking Introduces an approach to improve e-commerce product search by using dual-loss optimization to handle semantically relevant but intent-mismatched results. https://arxiv.org/abs/2410.15930 https://arxiv.org/abs/2410.15930" [X Link](https://x.com/_reachsumit/status/1848939507107971318) 2024-10-23T04:08Z [----] followers, [----] engagements "HyQE: Ranking Contexts with Hypothetical Query Embeddings Intuit presents a scalable context ranking framework that uses LLMs to generate hypothetical queries from retrieved contexts and ranks them. https://github.com/zwc662/hyqe https://arxiv.org/abs/2410.15262 https://github.com/zwc662/hyqe https://arxiv.org/abs/2410.15262" [X Link](https://x.com/_reachsumit/status/1848945365095211158) 2024-10-23T04:31Z [----] followers, [---] engagements "Class-RAG: Content Moderation with Retrieval Augmented Generation Meta GenAI Team introduces a flexible content moderation system that uses RAG to improve classification accuracy and adaptability. https://arxiv.org/abs/2410.14881 https://arxiv.org/abs/2410.14881" [X Link](https://x.com/_reachsumit/status/1848947324099395865) 2024-10-23T04:39Z [----] followers, [---] engagements "Bridging Search and Recommendation in Generative Retrieval: Does One Task Help the Other Spotify demonstrates how unifying search and recommendation tasks into a single LLM-based generative model can outperform specialized models. https://arxiv.org/abs/2410.16823 https://arxiv.org/abs/2410.16823" [X Link](https://x.com/_reachsumit/status/1848953215293002226) 2024-10-23T05:02Z [----] followers, [----] engagements "SouLLMate: An Application Enhancing Diverse Mental Health Support with Adaptive LLMs Prompt Engineering and RAG Techniques Presents an LLM-powered mental health support system achieving 80% accuracy in clinical assessments. https://github.com/QM378/SouLLMate https://arxiv.org/abs/2410.16322 https://github.com/QM378/SouLLMate https://arxiv.org/abs/2410.16322" [X Link](https://x.com/_reachsumit/status/1848955951665320433) 2024-10-23T05:13Z [----] followers, [---] engagements "DeCoRe: Decoding by Contrasting Retrieval Heads to Mitigate Hallucinations Reduces hallucinations in LLMs by contrasting outputs between normal and deliberately impaired versions of the model (via masked retrieval heads). https://github.com/aryopg/DeCoRe https://arxiv.org/abs/2410.18860 https://github.com/aryopg/DeCoRe https://arxiv.org/abs/2410.18860" [X Link](https://x.com/_reachsumit/status/1849667809665744932) 2024-10-25T04:22Z [----] followers, [----] engagements "Little Giants: Synthesizing High-Quality Embedding Data at Scale Trains small (8B) language models to generate high-quality synthetic data for text embeddings matching the performance of GPT-4-based approaches with less than 1/10 of the API calls. https://arxiv.org/abs/2410.18634 https://arxiv.org/abs/2410.18634" [X Link](https://x.com/_reachsumit/status/1849668705694908456) 2024-10-25T04:25Z [----] followers, [----] engagements "Quam: Adaptive Retrieval through Query Affinity Modelling Presents an improved adaptive retrieval approach that uses query affinity modeling to achieve better recall over standard re-ranking methods. https://github.com/Mandeep-Rathee/quam https://arxiv.org/abs/2410.20286 https://github.com/Mandeep-Rathee/quam https://arxiv.org/abs/2410.20286" [X Link](https://x.com/_reachsumit/status/1851127632316023079) 2024-10-29T05:03Z [----] followers, [---] engagements "Enhancing CTR prediction in Recommendation Domain with Search Query Representation Huawei introduces a framework that combines search query embeddings and collaborative filtering to boost CTR prediction by learning from users' search behavior patterns. https://arxiv.org/abs/2410.21487 https://arxiv.org/abs/2410.21487" [X Link](https://x.com/_reachsumit/status/1851465324383871319) 2024-10-30T03:25Z [----] followers, [---] engagements "Semantic Search Evaluation LinkedIn presents a method for evaluating content search systems by measuring semantic relevance between queries and results using LLMs introducing an "on-topic rate" metric for quality assessment. https://arxiv.org/abs/2410.21549 https://arxiv.org/abs/2410.21549" [X Link](https://x.com/_reachsumit/status/1851465520744394761) 2024-10-30T03:25Z [----] followers, [----] engagements "LLM-KT: A Versatile Framework for Knowledge Transfer from Large Language Models to Collaborative Filtering Enhances collaborative filtering models by injecting LLM-generated features into intermediate layers. https://github.com/a250/LLMRecSys_with_KnowledgeDistilation/tree/distil_framework https://arxiv.org/abs/2411.00556 https://github.com/a250/LLMRecSys_with_KnowledgeDistilation/tree/distil_framework https://arxiv.org/abs/2411.00556" [X Link](https://x.com/_reachsumit/status/1853288986082787421) 2024-11-04T04:11Z [----] followers, [----] engagements "Beyond Utility: Evaluating LLM as Recommender Proposes an evaluation framework for LLM-based recommender systems covering utility history length sensitivity candidate position bias generation quality and hallucinations. https://github.com/JiangDeccc/EvaLLMasRecommender https://arxiv.org/abs/2411.00331 https://github.com/JiangDeccc/EvaLLMasRecommender https://arxiv.org/abs/2411.00331" [X Link](https://x.com/_reachsumit/status/1853290573899370636) 2024-11-04T04:17Z [----] followers, [---] engagements "Rationale-Guided Retrieval Augmented Generation for Medical Question Answering Proposes a retrieval framework that enhances medical QA by filtering retrieved documents based on rationale perplexity. https://github.com/dmis-lab/RAG2 https://arxiv.org/abs/2411.00300 https://github.com/dmis-lab/RAG2 https://arxiv.org/abs/2411.00300" [X Link](https://x.com/_reachsumit/status/1853292230976643391) 2024-11-04T04:24Z [----] followers, [----] engagements "PSL: Rethinking and Improving Softmax Loss from Pairwise Perspective for Recommendation Introduces Pairwise Softmax Loss that provides tighter bounds for ranking metrics and better robustness against noise and dist shifts. https://github.com/Tiny-Snow/IR-Benchmark https://arxiv.org/abs/2411.00163 https://github.com/Tiny-Snow/IR-Benchmark https://arxiv.org/abs/2411.00163" [X Link](https://x.com/_reachsumit/status/1853292509352653038) 2024-11-04T04:25Z [----] followers, [----] engagements "JudgeRank: Leveraging Large Language Models for Reasoning-Intensive Reranking Introduces an agentic reranker that improves document retrieval by analyzing queries and documents through explicit reasoning steps. https://arxiv.org/abs/2411.00142 https://arxiv.org/abs/2411.00142" [X Link](https://x.com/_reachsumit/status/1853293536466330060) 2024-11-04T04:29Z [----] followers, [----] engagements "LLM4PR: Improving Post-Ranking in Search Engine with Large Language Models Kuaishou introduces the first LLM-based framework for post-ranking optimization in search engines leveraging a Query-Instructed Adapter and feature adaptation steps. https://arxiv.org/abs/2411.01178 https://arxiv.org/abs/2411.01178" [X Link](https://x.com/_reachsumit/status/1853669753203880256) 2024-11-05T05:24Z [----] followers, [---] engagements "MM-EMBED: Universal Multimodal Retrieval with Multimodal LLMs NVIDIA presents a versatile multimodal retriever that achieves SOTA on both multimodal and text retrieval tasks while maintaining strong text-to-text capabilities. https://huggingface.co/nvidia/MM-Embed https://arxiv.org/abs/2411.02571 https://huggingface.co/nvidia/MM-Embed https://arxiv.org/abs/2411.02571" [X Link](https://x.com/_reachsumit/status/1854017942167683407) 2024-11-06T04:28Z [----] followers, [----] engagements "Benchmarking Multimodal Retrieval Augmented Generation with Dynamic VQA Dataset and Self-Adaptive Planning Agent Alibaba introduces a self-adaptive planning agent for multimodal retrieval along with a new dataset. https://github.com/Alibaba-NLP/OmniSearch https://arxiv.org/abs/2411.02937 https://github.com/Alibaba-NLP/OmniSearch https://arxiv.org/abs/2411.02937" [X Link](https://x.com/_reachsumit/status/1854018318916841624) 2024-11-06T04:29Z [----] followers, [---] engagements "RAG-QA Arena: Evaluating Domain Robustness for Long-Form Retrieval-Augmented Question Answering Aamazon presents a new dataset with human-written long-form answers across [--] domains and proposes an evaluation framework. https://github.com/awslabs/rag-qa-arena https://www.amazon.science/publications/rag-qa-arena-evaluating-domain-robustness-for-long-form-retrieval-augmented-question-answering https://github.com/awslabs/rag-qa-arena https://www.amazon.science/publications/rag-qa-arena-evaluating-domain-robustness-for-long-form-retrieval-augmented-question-answering" [X Link](https://x.com/_reachsumit/status/1854019591779025148) 2024-11-06T04:34Z [----] followers, [----] engagements "HtmlRAG: HTML is Better Than Plain Text for Modeling Retrieved Knowledge in RAG Systems Leverages HTML structures instead of plain text for RAG with effective HTML cleaning and pruning strategies to maintain efficiency. https://github.com/plageon/HtmlRAG https://arxiv.org/abs/2411.02959 https://github.com/plageon/HtmlRAG https://arxiv.org/abs/2411.02959" [X Link](https://x.com/_reachsumit/status/1854019793986469962) 2024-11-06T04:35Z [----] followers, [---] engagements "RAGulator: Lightweight Out-of-Context Detectors for Grounded Text Generation Introduces a lightweight approach to detect out-of-context outputs in RAG systems using discriminative models trained on minimal resources. https://arxiv.org/abs/2411.03920 https://arxiv.org/abs/2411.03920" [X Link](https://x.com/_reachsumit/status/1854374738048827707) 2024-11-07T04:06Z [----] followers, [----] engagements "Long Context RAG Performance of Large Language Models Databricks analyzes [--] LLMs and reveals that only recent state-of-the-art models maintain consistent RAG accuracy above 64k tokens with most models' performance declining at longer contexts. https://arxiv.org/abs/2411.03538 https://arxiv.org/abs/2411.03538" [X Link](https://x.com/_reachsumit/status/1854375432046657679) 2024-11-07T04:08Z [----] followers, 12.5K engagements "Lightning IR: Straightforward Fine-tuning and Inference of Transformer-based Language Models for Information Retrieval Introduces a PyTorch Lightning-based framework for fine-tuning and inference of transformer models in IR. https://github.com/webis-de/lightning-ir https://arxiv.org/abs/2411.04677 https://github.com/webis-de/lightning-ir https://arxiv.org/abs/2411.04677" [X Link](https://x.com/_reachsumit/status/1854742473861673284) 2024-11-08T04:27Z [----] followers, [----] engagements "Best Practices for Distilling Large Language Models into BERT for Web Search Ranking Tencent introduces a distillation framework that effectively transfers LLM ranking capabilities to BERT models while dramatically reducing inference costs. https://arxiv.org/abs/2411.04539 https://arxiv.org/abs/2411.04539" [X Link](https://x.com/_reachsumit/status/1854742868080116029) 2024-11-08T04:28Z [----] followers, 10.6K engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@_reachsumit Sumit
Sumit posts on X about kuaishou, generative, alibaba, llm the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours.
Engagements: [-----] #
- [--] Week [------] +62%
- [--] Month [------] +17%
- [--] Months [-------] +166%
- [--] Year [-------] -54%
Mentions: [--] #
- [--] Month [--] -67%
- [--] Months [---] +62%
- [--] Year [---] +81%
Followers: [-----] #
- [--] Week [-----] +0.74%
- [--] Month [-----] +3%
- [--] Months [-----] +14%
- [--] Year [-----] +29%
CreatorRank: [---------] #
Social Influence
Social category influence technology brands social networks stocks finance fashion brands currencies cryptocurrencies
Social topic influence kuaishou #53, generative #280, alibaba #361, llm #408, tencent #175, bytedance, meituan #16, prediction, meta, realtime
Top accounts mentioned or mentioned by @jeremyphoward @bowenjin13 @xiaopengli @kvachai @youganglyu @junruwu4 @zeyuanmeng @louisackerman74 @genksn4 @yashardel @rayluthu @jinyeopsong @seungheondoh @seirasto @gsehgal1997 @guz @hschechtr @mufeili @jlwu002 @aksh555
Top assets mentioned Microsoft Corp. (MSFT) Alphabet Inc Class A (GOOGL) IBM (IBM) Spotify Technology (SPOT) Snap, Inc. (SNAP) Lossless (LSS) Intuit Inc. (INTU) Merge (MERGE) Frontier (FRONT) Fusion (FSN) Linear (LINA) Fetch (FET)
Top Social Posts
Top posts by engagements in the last [--] hours
"Revisiting Neural Retrieval on Accelerators Proposes a non-dot-product retrieval approach called "mixture of logits" (MoL) that models user-item similarity as an adaptive composition of elementary similarity functions. https://arxiv.org/abs/2306.04039 https://arxiv.org/abs/2306.04039"
X Link 2023-06-08T06:17Z [----] followers, [---] engagements
"Contrastive Multi-view Framework for Customer Lifetime Value Prediction Utilizes multiple heterogeneous regressors to extract complementary knowledge synthesizes predictions into a unified score & combines it with a purchase classifier to predict LTV https://arxiv.org/abs/2306.14400 https://arxiv.org/abs/2306.14400"
X Link 2023-06-27T03:51Z [----] followers, [--] engagements
"Structure Guided Multi-modal Pre-trained Transformer for Knowledge Graph Reasoning Leverages the structural information in multimodal knowledge graphs by adopting a graph structure encoder and a structure-guided fusion module"
X Link 2023-07-10T03:13Z [---] followers, [---] engagements
"Lexically-Accelerated Dense Retrieval Proposes a technique using lexical retrieval to seed and accelerate dense retrieval exploration via a document proximity graph. https://github.com/Georgetown-IR-Lab/ladr https://arxiv.org/abs/2307.16779 https://github.com/Georgetown-IR-Lab/ladr https://arxiv.org/abs/2307.16779"
X Link 2023-08-01T14:30Z [----] followers, [----] engagements
"NEON: Living Needs Prediction System in Meituan Proposes a system to predict living needs using spatiotemporal and behavioral features with multitask learning deployed and evaluated on Meituan"
X Link 2023-08-01T14:38Z [---] followers, [---] engagements
"SPM: Structured Pretraining and Matching Architectures for Relevance Modeling in Meituan Search Proposes a relevance matching architecture for ecommerce search queries and structured documents using multiple document fields and query-guided extractors"
X Link 2023-08-16T04:57Z [---] followers, [---] engagements
"I published Vol. [--] of "Top Information Retrieval Papers of the Week" on Substack"
X Link 2023-08-20T14:26Z [---] followers, [--] engagements
"Exploring the Integration Strategies of Retriever and Large Language Models Explores ways to improve open-domain question answering by better integrating retrieved passages with large language models and suggests four strategies. https://arxiv.org/abs/2308.12574 https://arxiv.org/abs/2308.12574"
X Link 2023-08-25T04:03Z [----] followers, [---] engagements
"Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations Combines recommender models and LLMs for versatile and interactive recommendation through efficient components like memory bus and demonstration-augmented planning"
X Link 2023-09-01T02:12Z [---] followers, [--] engagements
"Personalized Search Via Neural Contextual Semantic Relevance Ranking Proposes a personalized LTR framework that models relationships between user query context and document content using lexical and semantic representations to improve search relevance"
X Link 2023-09-12T03:32Z [---] followers, [---] engagements
"Retrieving Supporting Evidence for Generative Question Answering Explores using large language models to self-detect hallucinations in generated text by prompting the model to verify its responses against supporting evidence retrieved from a corpus"
X Link 2023-09-21T05:23Z [---] followers, [--] engagements
"Cluster Language Model for Improved E-Commerce Retrieval and Ranking: Leveraging Query Similarity and Fine-Tuning for Personalized Results Combines pre-trained LMs with query clustering to offer improved e-commerce retrieval and ranking performance"
X Link 2023-09-26T20:00Z [---] followers, [---] engagements
"RankVicuna: Zero-Shot Listwise Document Reranking with Open-Source Large Language Models Introduces an open-source listwise reranking approach using a 7B parameter language model for zero-shot document reranking"
X Link 2023-09-27T04:39Z [---] followers, [---] engagements
"The optimal way to utilize multimodal features is through end-to-end training of the recommender model and the video encoder. This is likely because E2E training can incorporate both raw video features and collaborative signals from user-item interactions"
X Link 2023-09-28T05:08Z [---] followers, [--] engagements
"Aligning the Capabilities of Large Language Models with the Context of Information Retrieval via Contrastive Feedback Proposes a framework called RLCF to align LLMs to improve their ability to generate specific and distinguishable responses"
X Link 2023-10-02T02:17Z [---] followers, [----] engagements
"Our Modern Recommendation Systems (MRS) team at Meta is hiring IC4/5/6 MLEs. Join us"
X Link 2023-10-07T00:42Z [---] followers, [----] engagements
"Lending Interaction Wings to Recommender Systems with Conversational Agents Proposes a plug-and-play framework that incorporates conversational agents into recsys by treating the recommender as a relevance estimator and the agent as a score checker"
X Link 2023-10-10T00:44Z [---] followers, [---] engagements
"Factual and Personalized Recommendations using Language Models and Reinforcement Learning Proposes a "personalized and compelling" language model for conversational recommender systems which uses reinforcement learning with AI feedback"
X Link 2023-10-11T06:06Z [---] followers, [---] engagements
"Retrieve Anything To Augment Large Language Models Proposes a unified embedding model trained to support diverse retrieval needs of large language models including knowledge memory examples and tools"
X Link 2023-10-12T01:52Z [---] followers, [---] engagements
"Found in the Middle: Permutation Self-Consistency Improves Listwise Ranking in Large Language Models Proposes generating multiple rankings from permutations of a prompt list and aggregates them to mitigate positional biases"
X Link 2023-10-12T03:17Z [---] followers, [---] engagements
"A Setwise Approach for Effective and Highly Efficient Zero-shot Ranking with Large Language Models Proposes a prompting approach for zero-shot document ranking with LLMs outperforming existing methods"
X Link 2023-10-18T05:21Z [---] followers, [---] engagements
"Chameleon: a Heterogeneous and Disaggregated Accelerator System for Retrieval-Augmented Language Models Proposes an accelerator system with specialized hardware for language model inference & vector search to efficiently serve diverse RALM configs"
X Link 2023-10-18T05:32Z [---] followers, [---] engagements
"Search-Adaptor: Text Embedding Customization for Information Retrieval Google Cloud AI proposes an adaptation method that improves text embeddings from LLMs (even API-based ones) for information retrieval by using a ranking loss and regularizers"
X Link 2023-10-18T05:42Z [---] followers, [----] engagements
"A Comprehensive Survey on Vector Database: Storage and Retrieval Technique Challenge Provides a review of vector database architectures approximate nearest neighbor search algorithms challenges and emerging directions involving LLMs"
X Link 2023-10-19T03:53Z [---] followers, [---] engagements
"Self-RAG: Learning to Retrieve Generate and Critique through Self-Reflection Proposes a framework that trains language models to retrieve knowledge on-demand and reflect on retrieved passages and their own generations"
X Link 2023-10-19T04:02Z [---] followers, [---] engagements
"Know Where to Go: Make LLM a Relevant Responsible and Trustworthy Searcher Proposes a generative retrieval framework with generator validator optimizer modules to improve the relevance responsibility & trustfulness of LLM-based retrieval systems"
X Link 2023-10-20T14:40Z [---] followers, [---] engagements
"Large Search Model: Redefining Search Stack in the Era of LLMs Microsoft proposes using a single large language model for all search tasks instead of many specialized models formulating tasks as text generation from prompts"
X Link 2023-10-26T19:06Z [---] followers, 13.8K engagements
"PaRaDe: Passage Ranking using Demonstrations with Large Language Models Google explores using difficulty-based selection to automatically choose effective demonstrations for improving LLM passage ranking and question generation"
X Link 2023-10-26T19:12Z [---] followers, [---] engagements
"One Model for All: Large Language Models are Domain-Agnostic Recommendation Systems Ant Group proposes using LLMs for multi-domain sequential recommendation by representing users and items with concatenated item titles to leverage pretrained knowledge"
X Link 2023-10-26T19:21Z [---] followers, [----] engagements
"Representation Learning with Large Language Models for Recommendation Proposes a framework to enhance existing recommender systems by aligning semantic representations from LLMs with collaborative representations"
X Link 2023-10-26T19:49Z [---] followers, [---] engagements
"Open-source Large Language Models are Strong Zero-shot Query Likelihood Models for Document Ranking Investigates the zero-shot ranking effectiveness of LLMs as QLMs and proposes to integrate them with a hybrid retriever"
X Link 2023-10-27T04:57Z [---] followers, [---] engagements
"CoLLM: Integrating Collaborative Embeddings into Large Language Models for Recommendation Proposes to enhance LLM recommenders by incorporating collaborative information from traditional models thru embedding space alignment"
X Link 2023-11-01T01:27Z [---] followers, [---] engagements
"MILL: Mutual Verification with Large Language Models for Zero-Shot Query Expansion Proposes a framework for query expansion which generates documents through q-q-d prompts and mutually filters retrieved and generated documents to complement each other"
X Link 2023-11-01T01:53Z [---] followers, [----] engagements
"LLMRec: Large Language Models with Graph Augmentation for Recommendation Proposes a framework that LLMs to augment recommender systems by reinforcing user-item edges enhancing item attributes and profiling users"
X Link 2023-11-02T02:12Z [---] followers, [----] engagements
"Collaborative Large Language Model for Recommender Systems LinkedIn proposes a generative recommender system that tightly integrates the ID paradigm and LLM paradigm using extended vocabulary novel pretraining and finetuning"
X Link 2023-11-03T14:38Z [---] followers, [----] engagements
"Instruction Distillation Makes Large Language Models Efficient Zero-shot Rankers Distills complex pairwise ranking instructions into simpler pointwise instructions to improve the effectiveness of LLMs for zero-shot ranking"
X Link 2023-11-06T06:22Z [---] followers, 10.5K engagements
"LlamaRec: Two-Stage Recommendation using Large Language Models for Ranking Proposes a two-stage recommendation framework that uses a sequential model for efficient candidate retrieval and Llama [--] for effective ranking"
X Link 2023-11-07T04:07Z [---] followers, [----] engagements
"Knowledge-Augmented Large Language Models for Personalized Contextual Query Suggestion Microsoft Research presents a method to personalize LLMs for search via entity-based user knowledge stores derived from logs"
X Link 2023-11-14T07:03Z [---] followers, 19.1K engagements
"Text Retrieval with Multi-Stage Re-Ranking Models Hitachi roposes a 3-stage retrieval model using BM25 language model and model ensemble/larger model to improve accuracy while reducing search delays"
X Link 2023-11-15T07:10Z [---] followers, [---] engagements
"Chain-of-Note: Enhancing Robustness in Retrieval-Augmented Language Models Tencent AI Lab proposes Chain-of-Noting to improve retrieval-augmented language models through robustness to irrelevant retrieved documents and unknown queries"
X Link 2023-11-16T06:04Z [---] followers, [---] engagements
"@jeremyphoward Congratulations Pat Cummins was absolutely brilliant tonight and a nightmare to face as the opposing team's captain. I'll hold a grudge against him for the rest of his cricketing career. ðŸ˜ðŸ˜ðŸ˜‚"
X Link 2023-11-19T22:11Z [---] followers, [---] engagements
"RankingGPT: Empowering Large Language Models in Text Ranking with Progressive Enhancement Alibaba trains LLMs for text ranking through weak supervision & fine-tuning to align objectives to achieve state-of-the-art performance"
X Link 2023-11-29T03:56Z [---] followers, [---] engagements
"ControlRec: Bridging the Semantic Gap between Language Model and Personalized Recommendation Meituan proposes to align user/item IDs with language semantics in recommendations via contrastive learning for improved generalization"
X Link 2023-11-29T04:03Z [---] followers, [---] engagements
"Event-driven Real-time Retrieval in Web Search Tencent proposes Event-driven Real-time Retrieval an approach that expands queries with breaking events to improve real-time search through cross-attention and multi-task training"
X Link 2023-12-04T06:26Z [---] followers, [---] engagements
"Lookahead: An Inference Acceleration Framework for Large Language Model with Lossless Generation Accuracy Proposes Lookahead a multi-branch decoding framework using trie-based retrieval for accelerating inference of LLMs by overcoming I/O constraints"
X Link 2023-12-22T06:50Z [---] followers, [---] engagements
"GUITAR: Gradient Pruning toward Fast Neural Ranking Baidu Research proposes a fast graph search method for neural ranking that uses gradient-based candidate pruning and an adaptive heuristic to minimize expensive neural network computations"
X Link 2023-12-29T04:54Z [---] followers, [---] engagements
"Searching fast and slow through product catalogs Microsoft presents a fast and accurate SKU search system for CRMs combining Trie-based suggestions TF-IDF retrieval and language model embeddings outperforming existing systems"
X Link 2024-01-02T02:42Z [---] followers, 16K engagements
"A case study of Generative AI in MSX Sales Copilot: Improving seller productivity with a real-time question-answering system for content recommendation Microsoft presents a real-time QA system that matches seller queries to sales documentation"
X Link 2024-01-11T03:56Z [---] followers, [---] engagements
"ChatQA: Building GPT-4 Level Conversational QA Models Nvidia introduces a family of conversational QA models matching GPT-4 performance using efficient training methods like two-stage instruction tuning and improved context retrieval"
X Link 2024-01-19T02:19Z [--] followers, [---] engagements
"In-context Learning with Retrieved Demonstrations for Language Models: A Survey Google Research offers a comprehensive analysis of retrieval-based ICL highlighting key innovations and future paths to enhance demonstration relevance and diversity"
X Link 2024-01-23T03:48Z [---] followers, [----] engagements
"Enhancing Recommendation Diversity by Re-ranking with Large Language Models Shows that prompting LLMs to rerank recommendations for diversity shows promise but currently underperforms traditional diversification techniques"
X Link 2024-01-23T05:14Z [---] followers, [---] engagements
"Multimodal Query Suggestion with Multi-Agent Reinforcement Learning from Human Feedback Huawei presents a new task to generate query suggestions from user images rather than just text to better capture intent and diversity"
X Link 2024-02-08T06:05Z [---] followers, [---] engagements
"RA-Rec: An Efficient ID Representation Alignment Framework for LLM-based Recommendation Huawei presents a framework that aligns ID embeddings with LLMs for recommendation systems to leverage the strengths of both approaches"
X Link 2024-02-08T06:20Z [---] followers, [---] engagements
"CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models Presents a LLM-powered plugin for Slack facilitating collaborative search within conversations by understanding context and providing summaries"
X Link 2024-02-12T02:16Z [---] followers, [----] engagements
"Search Intention Network for Personalized Query Auto-Completion in E-Commerce Alibaba proposes a framework for query auto-completion that models a user's current intention from the prefix and historical intentions from behavior sequences"
X Link 2024-03-06T07:29Z [----] followers, [---] engagements
"Aligning Large Language Models for Controllable Recommendations Microsoft proposes a two-stage supervised and RL approach to enhance LLMs' ability to follow instructions and reduce errors in recommender systems"
X Link 2024-03-11T04:57Z [----] followers, [---] engagements
"MetaSplit: Meta-Split Network for Limited-Stock Product Recommendation Alibaba proposes addressing the challenges of recommending limited-stock products in C2C e-commerce platforms by employing differentiated modeling approaches based on stock levels"
X Link 2024-03-12T03:43Z [----] followers, [---] engagements
"Detecting Hallucination and Coverage Errors in Retrieval Augmented Generation for Controversial Topics Google proposes using RAG with LLM-based classifiers trained on synthetic errors to handle controversial topics"
X Link 2024-03-15T02:06Z [----] followers, [---] engagements
"SelfIE: Self-Interpretation of Large Language Model Embeddings Introduces a framework that enables LLMs to interpret their own hidden embeddings in natural language revealing their internal reasoning processes. https://selfie.cs.columbia.edu/ https://arxiv.org/abs/2403.10949 https://selfie.cs.columbia.edu/ https://arxiv.org/abs/2403.10949"
X Link 2024-03-19T03:46Z [----] followers, [---] engagements
"AlphaFin: Benchmarking Financial Analysis with Retrieval-Augmented Stock-Chain Framework Alibaba proposes a framework that integrates LLMs with a real-time financial database to perform stock trend prediction and financial QA"
X Link 2024-03-20T04:20Z [----] followers, [---] engagements
"CoLLEGe: Concept Embedding Generation for Large Language Models Presents a meta-learning framework that enables LLMs to quickly acquire new concepts from few examples by generating embeddings"
X Link 2024-03-25T01:55Z [----] followers, [----] engagements
"MA4DIV: Multi-Agent Reinforcement Learning for Search Result Diversification Models search result diversification as a cooperative task among document agents directly optimizing diversity metrics while improving efficiency over existing methods. https://arxiv.org/abs/2403.17421 https://arxiv.org/abs/2403.17421"
X Link 2024-03-27T06:34Z [----] followers, [---] engagements
"Make Large Language Model a Better Ranker Aligns large language models with listwise ranking objectives for recommender systems using soft lambda loss and permutation-sensitive learning to address computational cost and position bias challenges. https://arxiv.org/abs/2403.19181 https://arxiv.org/abs/2403.19181"
X Link 2024-03-29T04:26Z [----] followers, [---] engagements
"FT2Ra: A Fine-Tuning-Inspired Approach to Retrieval-Augmented Code Completion Introduces a retrieval-based approach that mimics fine-tuning by iteratively augmenting a code pre-trained model's predictions using retrieved delta logits. https://arxiv.org/abs/2404.01554 https://arxiv.org/abs/2404.01554"
X Link 2024-04-03T06:33Z [----] followers, [---] engagements
"Event-enhanced Retrieval in Real-time Search Tencent tackles the "semantic drift" problem in real-time search by incorporating a generative decoder to extract event-focused representations. https://github.com/open-event-hub/Event-enhanced_Retrieval https://arxiv.org/abs/2404.05989 https://github.com/open-event-hub/Event-enhanced_Retrieval https://arxiv.org/abs/2404.05989"
X Link 2024-04-11T05:21Z [----] followers, [---] engagements
"Graph Chain-of-Thought: Augmenting Large Language Models by Reasoning on Graphs Introduces a benchmark for evaluating how LLMs can leverage KGs and proposes a framework that enables iterative reasoning on graphs. https://arxiv.org/abs/2404.07103 https://github.com/PeterGriffinJin/Graph-CoT https://arxiv.org/abs/2404.07103 https://github.com/PeterGriffinJin/Graph-CoT"
X Link 2024-04-11T05:36Z [----] followers, [---] engagements
"Reducing hallucination in structured outputs via Retrieval-Augmented Generation ServiceNow proposes to reduce hallucinations and improve generalization in a system that converts natural language requirements into structured workflow representations. https://arxiv.org/abs/2404.08189 https://arxiv.org/abs/2404.08189"
X Link 2024-04-15T03:12Z [----] followers, [----] engagements
"Navigating the Evaluation Funnel to Optimize Iteration Speed for Recommender Systems Spotify presents a framework for evaluating recommendation systems by decomposing the definition of success and combining offline and online evaluation methods. https://arxiv.org/abs/2404.08671 https://arxiv.org/abs/2404.08671"
X Link 2024-04-16T05:38Z [----] followers, [---] engagements
"Retrieval Augmented Generation for Domain-specific Question Answering Adobe proposes a framework for developing an in-house QA system featuring a large QA database a retriever trained on domain data and user behavior and a finetuned language model. https://arxiv.org/abs/2404.14760 https://arxiv.org/abs/2404.14760"
X Link 2024-04-24T02:14Z [----] followers, [----] engagements
"Hi-Gen: Generative Retrieval For Large-Scale Personalized E-commerce Search Alibaba encodes semantic information in document ID generation & exploits positional information during decoding to enable efficient large-scale personalized E-commerce search. https://arxiv.org/abs/2404.15675 https://arxiv.org/abs/2404.15675"
X Link 2024-04-25T06:51Z [----] followers, [----] engagements
"OmniSearchSage: Multi-Task Multi-Entity Embeddings for Pinterest Search Pinterest jointly learns unified query pin and product embeddings using enriched entity representations. https://github.com/pinterest/atg-research/tree/main/omnisearchsage https://arxiv.org/abs/2404.16260 https://github.com/pinterest/atg-research/tree/main/omnisearchsage https://arxiv.org/abs/2404.16260"
X Link 2024-04-26T01:28Z [----] followers, [----] engagements
"InspectorRAGet: An Introspection Platform for RAG Evaluation Enables comprehensive evaluation of retrieval-augmented language models through aggregate and instance-level analysis mixed metrics and annotator assessment. https://github.com/IBM/InspectorRAGet https://arxiv.org/abs/2404.17347 https://github.com/IBM/InspectorRAGet https://arxiv.org/abs/2404.17347"
X Link 2024-04-29T03:28Z [----] followers, [----] engagements
"Interest Clock: Time Perception in Real-Time Streaming Recommendation System Douyin/Bytedance encodes users' time-aware preferences into personalized clock embeddings and employs Gaussian smoothing to capture dynamic user preferences over time. https://arxiv.org/abs/2404.19357 https://arxiv.org/abs/2404.19357"
X Link 2024-05-01T01:36Z [----] followers, [---] engagements
""In-Context Learning" or: How I learned to stop worrying and love "Applied Information Retrieval" Draws an analogy between ICL for LLMs and IR proposes IR techniques to improve few-shot example selection and enhance downstream genAI tasks. https://arxiv.org/abs/2405.01116 https://arxiv.org/abs/2405.01116"
X Link 2024-05-03T02:28Z [----] followers, [----] engagements
"Semi-Parametric Retrieval via Binary Token Index Utilizes efficient binary token indexing alongside embedding-based indexing achieving superior retrieval accuracy while reducing indexing time and storage requirements https://github.com/jzhoubu/VDR https://arxiv.org/abs/2405.01924 https://github.com/jzhoubu/VDR https://arxiv.org/abs/2405.01924"
X Link 2024-05-06T06:21Z [----] followers, [---] engagements
"Lifelong Knowledge Editing for LLMs with Retrieval-Augmented Continuous Prompt Learning Alibaba enables lifelong model editing for LLMs by converting knowledge into continuous prompts and employing dynamic prompt retrieval. https://arxiv.org/abs/2405.03279 https://arxiv.org/abs/2405.03279"
X Link 2024-05-07T03:46Z [----] followers, [---] engagements
"R4: Reinforced Retriever-Reorder-Responder for Retrieval-Augmented Large Language Models Alibaba improves retrieval-augmented LLMs by learning optimal document orderings and enhancing document representations via RL and adversarial techniques. https://arxiv.org/abs/2405.02659 https://arxiv.org/abs/2405.02659"
X Link 2024-05-07T04:14Z [----] followers, [---] engagements
"FlashBack:Efficient Retrieval-Augmented Language Modeling for Long Context Inference Appends retrieved documents to improve inference efficiency while maintaining performance through tunable Marking Tokens and LoRA fine-tuning. https://arxiv.org/abs/2405.04065 https://arxiv.org/abs/2405.04065"
X Link 2024-05-08T04:48Z [----] followers, [---] engagements
"Knowledge Adaptation from Large Language Model to Recommendation for Practical Industrial Application Kuaishou employs a frozen LLM for content embedding extraction and connects the open-world and collaborative knowledge domains through an SSL task. https://arxiv.org/abs/2405.03988 https://arxiv.org/abs/2405.03988"
X Link 2024-05-08T04:54Z [----] followers, [---] engagements
"Positional encoding is not the same as context: A study on positional encoding for Sequential recommendation Huawei analyzes positional encodings in transformer-based sequential recommendation systems proposing new encodings. https://github.com/researcher1741/Position_encoding_SRS https://arxiv.org/abs/2405.10436 https://github.com/researcher1741/Position_encoding_SRS https://arxiv.org/abs/2405.10436"
X Link 2024-05-20T02:51Z [----] followers, [----] engagements
"Empowering Small-Scale Knowledge Graphs: A Strategy of Leveraging General-Purpose Knowledge Graphs for Enriched Embeddings Proposes a framework to enrich small domain-specific knowledge graphs with large general-purpose KGs. https://github.com/graphml-lab-pwr/empowering-small-scale-kg https://arxiv.org/abs/2405.10745 https://github.com/graphml-lab-pwr/empowering-small-scale-kg https://arxiv.org/abs/2405.10745"
X Link 2024-05-20T02:58Z [----] followers, [----] engagements
"From Sora What We Can See: A Survey of Text-to-Video Generation Reviews the text-to-video generation domain centered around OpenAI's Sora model identifying remaining challenges and proposing future research directions. https://github.com/soraw-ai/Awesome-Text-to-Video-Generation https://arxiv.org/abs/2405.10674 https://github.com/soraw-ai/Awesome-Text-to-Video-Generation https://arxiv.org/abs/2405.10674"
X Link 2024-05-20T03:03Z [----] followers, [----] engagements
"Diversifying by Intent in Recommender Systems Google proposes a probabilistic intent-based diversification framework for recommender systems that incorporates higher-level user intents to optimize long-term user experience. https://arxiv.org/abs/2405.12327 https://arxiv.org/abs/2405.12327"
X Link 2024-05-22T02:22Z [----] followers, [---] engagements
"NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models Nvidia introduces an embedding model with architectural improvements like latent attention pooling and a 2-stage contrastive instruction-tuning. https://huggingface.co/nvidia/NV-Embed-v1 https://arxiv.org/abs/2405.17428 https://huggingface.co/nvidia/NV-Embed-v1 https://arxiv.org/abs/2405.17428"
X Link 2024-05-28T03:57Z [----] followers, [----] engagements
"Generative Query Reformulation Using Ensemble Prompting Document Fusion and Relevance Feedback Proposes ensemble-based prompting techniques that leverage paraphrastic instructions to generate diverse query reformulations. https://arxiv.org/abs/2405.17658 https://arxiv.org/abs/2405.17658"
X Link 2024-05-29T04:41Z [----] followers, [---] engagements
"RAG Does Not Work for Enterprises Explores the challenges and requirements for implementing RAG in enterprises proposing potential solutions like semantic search and hybrid queries and an evaluation framework to validate enterprise-grade RAG solutions https://arxiv.org/abs/2406.04369 https://arxiv.org/abs/2406.04369"
X Link 2024-06-10T03:30Z [----] followers, 107.1K engagements
"Async Learned User Embeddings for Ads Delivery Optimization Meta learns high-fidelity user embeddings from multimodal user activities using a Transformer-like model and graph learning enabling effective ad retrieval. https://arxiv.org/abs/2406.05898 https://arxiv.org/abs/2406.05898"
X Link 2024-06-11T04:17Z [----] followers, [---] engagements
"Optimizing Novelty of Top-k Recommendations using Large Language Models and Reinforcement Learning Microsoft optimizes novelty in large-scale recommendation systems using LLMs for feedback and a reformulated state-action space to improve efficiency. https://arxiv.org/abs/2406.14169 https://arxiv.org/abs/2406.14169"
X Link 2024-06-21T05:46Z [----] followers, [---] engagements
"Do Not Wait: Learning Re-Ranking Model Without User Feedback At Serving Time in E-Commerce Generates transient request-specific model modifications using a surrogate model enabling immediate adaptation without relying on delayed user feedback. https://arxiv.org/abs/2406.14004 https://arxiv.org/abs/2406.14004"
X Link 2024-06-21T06:12Z [----] followers, [---] engagements
"This post concludes the two-part series on the evolution of the MTL-based video recommender system. It describes disentanglement strategies to enable effective and efficient learning of inter-task relationships. https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/ https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/"
X Link 2024-06-22T17:34Z [----] followers, [---] engagements
"Additionally the post emphasizes real-world solutions from Kuaishou Tencent YouTube Facebook and Amazon Prime Video to address different biases and shares tips from various published research works. https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/ https://blog.reachsumit.com/posts/2024/06/multi-task-video-recsys-p2/"
X Link 2024-06-22T17:34Z [----] followers, [---] engagements
"FIRST: Faster Improved Listwise Reranking with Single Token Decoding Proposes an LLM reranking method that uses only the first token's logits to rank passages improving efficiency by 50% while maintaining performance. https://github.com/gangiswag/llm-reranker https://arxiv.org/abs/2406.15657 https://github.com/gangiswag/llm-reranker https://arxiv.org/abs/2406.15657"
X Link 2024-06-25T04:06Z [----] followers, [----] engagements
"T-FREE: Tokenizer-Free Generative LLMs via Sparse Representations for Memory-Efficient Embeddings Presents a tokenizer-free approach for LLMs that directly embeds words using character triplets showing competitive performance https://github.com/Aleph-Alpha/trigrams https://arxiv.org/abs/2406.19223 https://github.com/Aleph-Alpha/trigrams https://arxiv.org/abs/2406.19223"
X Link 2024-06-28T03:41Z [----] followers, [---] engagements
"Summary of a Haystack: A Challenge to Long-Context LLMs and RAG Systems Salesforce presents a new benchmark for evaluating long-context language models and RAG systems requiring precise summarization and citation of insights. https://github.com/salesforce/summary-of-a-haystack https://arxiv.org/abs/2407.01370 https://github.com/salesforce/summary-of-a-haystack https://arxiv.org/abs/2407.01370"
X Link 2024-07-02T15:59Z [----] followers, [----] engagements
"RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs NVIDIA introduces an instruction fine-tuning framework that enables a single LLM to perform both context ranking and answer generation in RAG. https://arxiv.org/abs/2407.02485 https://arxiv.org/abs/2407.02485"
X Link 2024-07-03T15:25Z [----] followers, 20.2K engagements
"MemoCRS: Memory-enhanced Sequential Conversational Recommender Systems with Large Language Models Huawei uses memory-enhanced LLMs to capture user preference continuity across dialogue sessions. https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/memocrs https://arxiv.org/abs/2407.04960 https://github.com/mindspore-lab/models/tree/master/research/huawei-noah/memocrs https://arxiv.org/abs/2407.04960"
X Link 2024-07-09T05:15Z [----] followers, [---] engagements
"Merge Ensemble and Cooperate A Survey on Collaborative Strategies in the Era of Large Language Models Examines collaboration strategies for LLMs categorizing them into Merging Ensemble and Cooperation approaches and provides reviews of each. https://arxiv.org/abs/2407.06089 https://arxiv.org/abs/2407.06089"
X Link 2024-07-09T05:23Z [----] followers, [---] engagements
"RAG vs. Long Context: Examining Frontier Large Language Models for Environmental Review Document Comprehension Introduces a benchmark for evaluating LLMs' performance on NEPA documents finding that RAG-powered models outperform long-context LLMs. https://arxiv.org/abs/2407.07321 https://arxiv.org/abs/2407.07321"
X Link 2024-07-11T02:22Z [----] followers, [----] engagements
"Deep Bag-of-Words Model: An Efficient and Interpretable Relevance Architecture for Chinese E-Commerce Alibaba combines deep learning with sparse Bag-of-Words representation improving performance interpretability and efficiency over dense models. https://arxiv.org/abs/2407.09395 https://arxiv.org/abs/2407.09395"
X Link 2024-07-15T02:18Z [----] followers, [----] engagements
"ChatQA 2: Bridging the Gap to Proprietary LLMs in Long Context and RAG Capabilities NVIDIA extends Llama3-70B to rival GPT-4-Turbo in long-context understanding and RAG achieving comparable performance on various tasks. https://arxiv.org/abs/2407.14482 https://arxiv.org/abs/2407.14482"
X Link 2024-07-22T02:29Z [----] followers, [----] engagements
"NV-Retriever: Improving text embedding models with effective hard-negative mining NVIDIA presents positive-aware hard-negative mining for text embedding models and a model that topped the MTEB Retrieval benchmark in July'24. https://huggingface.co/nvidia/NV-Retriever-v1 https://arxiv.org/abs/2407.15831 https://huggingface.co/nvidia/NV-Retriever-v1 https://arxiv.org/abs/2407.15831"
X Link 2024-07-23T05:39Z [----] followers, [----] engagements
"Efficient Retrieval with Learned Similarities Introduces Mixture-of-Logits (MoL) as a universal approximator for learned similarity functions in retrieval tasks proposing efficient techniques for approximate top-K retrieval. https://arxiv.org/abs/2407.15462 https://arxiv.org/abs/2407.15462"
X Link 2024-07-23T05:44Z [----] followers, [----] engagements
"TWIN V2: Scaling Ultra-Long User Behavior Sequence Modeling for Enhanced CTR Prediction at Kuaishou Kuaishou introduces a method to extend user behavior modeling to lifecycle scale (up to [---] interactions) for CTR prediction in recommendation systems. https://arxiv.org/abs/2407.16357 https://arxiv.org/abs/2407.16357"
X Link 2024-07-24T02:10Z [----] followers, [---] engagements
"RazorAttention: Efficient KV Cache Compression Through Retrieval Heads Huawei proposes a KV cache compression technique compatible with FlashAttention for LLMs that reduces cache size by over 70% without significant performance loss. https://arxiv.org/abs/2407.15891 https://arxiv.org/abs/2407.15891"
X Link 2024-07-24T02:15Z [----] followers, [---] engagements
"Improving Retrieval Augmented Language Model with Self-Reasoning Baidu enhances RALMs by using LLM-generated reasoning trajectories improving reliability and traceability rivaling GPT-4's performance on QA tasks using minimal training data. https://arxiv.org/abs/2407.19813 https://arxiv.org/abs/2407.19813"
X Link 2024-07-30T03:48Z [----] followers, [----] engagements
"Enhancing Taobao Display Advertising with Multimodal Representations: Challenges Approaches and Insights Alibaba presents a two-phase framework for integrating multimodal data into large-scale recommendation systems. https://arxiv.org/abs/2407.19467 https://arxiv.org/abs/2407.19467"
X Link 2024-07-30T04:35Z [----] followers, [---] engagements
"Breaking the Hourglass Phenomenon of Residual Quantization: Enhancing the Upper Bound of Generative Retrieval Identifies the "Hourglass" phenomenon in generative retrieval proposing solutions to mitigate token concentration in intermediate layers. https://arxiv.org/abs/2407.21488 https://arxiv.org/abs/2407.21488"
X Link 2024-08-01T02:23Z [----] followers, [---] engagements
"MLLM Is a Strong Reranker: Advancing Multimodal Retrieval-augmented Generation via Knowledge-enhanced Reranking and Noise-injected Training Presents a framework that addresses the multi-granularity noisy correspondence problem https://github.com/IDEA-FinAI/RagLLaVA https://arxiv.org/abs/2407.21439 https://github.com/IDEA-FinAI/RagLLaVA https://arxiv.org/abs/2407.21439"
X Link 2024-08-01T02:27Z [----] followers, [----] engagements
"Simple but Efficient: A Multi-Scenario Nearline Retrieval Framework for Recommendation on Taobao Alibaba introduces a framework that incorporates ranking results from various scenarios into the matching stage demonstrating a 5% increase in transactions https://arxiv.org/abs/2408.00247 https://arxiv.org/abs/2408.00247"
X Link 2024-08-02T02:38Z [----] followers, [---] engagements
"RECE: Reduced Cross-Entropy Loss for Large-Catalogue Sequential Recommenders Presents a loss function that approximates CE loss for sequential recommenders significantly reducing GPU memory usage while maintaining performance https://github.com/dalibra/RECE https://arxiv.org/abs/2408.02354 https://github.com/dalibra/RECE https://arxiv.org/abs/2408.02354"
X Link 2024-08-06T04:47Z [----] followers, [---] engagements
"A Real-Time Adaptive Multi-Stream GPU System for Online Approximate Nearest Neighborhood Search Xiaohongshu presents a GPU-based system for real-time Approximate Nearest Neighbor Search featuring dynamic vector insertion and multi-stream execution. https://arxiv.org/abs/2408.02937 https://arxiv.org/abs/2408.02937"
X Link 2024-08-07T02:04Z [----] followers, [----] engagements
"Relevance meets Diversity: A User-Centric Framework for Knowledge Exploration through Recommendations Introduces a user-centric recommender system that maximizes knowledge acquisition by balancing relevance and diversity. https://github.com/EricaCoppolillo/EXPLORE https://arxiv.org/abs/2408.03772 https://github.com/EricaCoppolillo/EXPLORE https://arxiv.org/abs/2408.03772"
X Link 2024-08-08T02:40Z [----] followers, [----] engagements
"Advancing Re-Ranking with Multimodal Fusion and Target-Oriented Auxiliary Tasks in E-Commerce Search JD.com presents a new e-commerce search re-ranking model that integrates multimodal information using attention-based fusion and auxiliary tasks. https://arxiv.org/abs/2408.05751 https://arxiv.org/abs/2408.05751"
X Link 2024-08-13T04:16Z [----] followers, [---] engagements
"HoME: Hierarchy of Multi-Gate Experts for Multi-Task Learning at Kuaishou Kuaishou presents a multi-task learning framework for short-video recommendations that addresses expert collapse degradation and underfitting in MoE systems. https://arxiv.org/abs/2408.05430 https://arxiv.org/abs/2408.05430"
X Link 2024-08-13T04:25Z [----] followers, [---] engagements
"Mamba Retriever: Utilizing Mamba for Effective and Efficient Dense Retrieval Presents a dense retrieval model based on the Mamba architecture offering superior efficiency for long-text retrieval due to its linear time scaling https://github.com/41924076/MambaRetriever https://arxiv.org/abs/2408.08066 https://github.com/41924076/MambaRetriever https://arxiv.org/abs/2408.08066"
X Link 2024-08-16T02:44Z [----] followers, [----] engagements
"ColBERT's MASK-based Query Augmentation: Effects of Quadrupling the Query Input Length While query effectiveness is optimized at a specific MASK token count the model remains robust even with significantly more tokens than it was trained on. https://arxiv.org/abs/2408.13672 https://arxiv.org/abs/2408.13672"
X Link 2024-08-27T04:45Z [----] followers, [----] engagements
"LRP4RAG: Detecting Hallucinations in Retrieval-Augmented Generation via Layer-wise Relevance Propagation Analyzes relevance between input and output and uses Layer-wise Relevance Propagation to detect hallucinations in RAG systems. https://arxiv.org/abs/2408.15533 https://arxiv.org/abs/2408.15533"
X Link 2024-08-29T02:44Z [----] followers, [---] engagements
"CRAFT Your Dataset: Task-Specific Synthetic Dataset Generation Through Corpus Retrieval and Augmentation Presents a method for generating task-specific synthetic datasets using user-provided few-shot examples. https://github.com/ziegler-ingo/CRAFT https://arxiv.org/abs/2409.02098 https://github.com/ziegler-ingo/CRAFT https://arxiv.org/abs/2409.02098"
X Link 2024-09-04T05:16Z [----] followers, [----] engagements
"Genetic Approach to Mitigate Hallucination in Generative IR Introduces a method using a genetic algorithm with a balanced fitness function to reduce hallucinations in generative language models. https://github.com/Georgetown-IR-Lab/GAuGE https://arxiv.org/abs/2409.00085 https://github.com/Georgetown-IR-Lab/GAuGE https://arxiv.org/abs/2409.00085"
X Link 2024-09-04T05:49Z [----] followers, [---] engagements
"Building a Scalable Effective and Steerable Search and Ranking Platform Zalando introduces a scalable real-time e-commerce ranking platform using transformer-based models to personalize product recommendations across various use cases. https://arxiv.org/abs/2409.02856 https://arxiv.org/abs/2409.02856"
X Link 2024-09-05T03:06Z [----] followers, [----] engagements
"RouterRetriever: Exploring the Benefits of Routing over Multiple Expert Embedding Models Introduces a flexible multi-expert approach to information retrieval that routes queries to domain-specific experts. https://github.com/amy-hyunji/RouterRetriever https://arxiv.org/abs/2409.02685 https://github.com/amy-hyunji/RouterRetriever https://arxiv.org/abs/2409.02685"
X Link 2024-09-05T03:13Z [----] followers, [----] engagements
"Diversify-verify-adapt: Efficient and Robust Retrieval-Augmented Ambiguous Question Answering Proposes a framework for handling ambiguous questions in RAG system using retrieval diversification and adaptive generation to improve accuracy and efficiency https://arxiv.org/abs/2409.02361 https://arxiv.org/abs/2409.02361"
X Link 2024-09-05T03:17Z [----] followers, [---] engagements
"MOBIUS: Towards the Next Generation of Query-Ad Matching in Baidu's Sponsored Search Baidu introduces a unified query-ad matching system that considers both relevance and commercial performance utilizing active learning and efficient ANN techniques. https://arxiv.org/abs/2409.03449 https://arxiv.org/abs/2409.03449"
X Link 2024-09-06T03:52Z [----] followers, [---] engagements
"Large Language Model-Based Agents for Software Engineering: A Survey Analyzes [---] papers on LLM-based agents in Software Engineering examining their applications across various development tasks and agent designs. https://github.com/FudanSELab/Agent4SE-Paper-List https://arxiv.org/abs/2409.02977 https://github.com/FudanSELab/Agent4SE-Paper-List https://arxiv.org/abs/2409.02977"
X Link 2024-09-06T03:55Z [----] followers, [----] engagements
"OneGen: Efficient One-Pass Unified Generation and Retrieval for LLMs Presents a framework that enables LLMs to perform both generation and retrieval in a single forward pass. https://github.com/zjunlp/OneGen https://arxiv.org/abs/2409.05152 https://github.com/zjunlp/OneGen https://arxiv.org/abs/2409.05152"
X Link 2024-09-10T04:41Z [----] followers, [----] engagements
"Late Chunking: Contextual Chunk Embeddings Using Long-Context Embedding Models Jina AI presents a technique that improves text embeddings for retrieval tasks by encoding entire documents before splitting them. https://github.com/jina-ai/late-chunking https://arxiv.org/abs/2409.04701 https://github.com/jina-ai/late-chunking https://arxiv.org/abs/2409.04701"
X Link 2024-09-10T04:50Z [----] followers, [----] engagements
"LexBoost: Improving Lexical Document Retrieval with Nearest Neighbors Uses pre-computed document neighborhoods to boost relevance scores and enhance lexical search. https://github.com/Georgetown-IR-Lab/LexBoost https://arxiv.org/abs/2409.05882 https://github.com/Georgetown-IR-Lab/LexBoost https://arxiv.org/abs/2409.05882"
X Link 2024-09-11T02:50Z [----] followers, [----] engagements
"RePlay: a Recommendation Framework for Experimentation and Production Use Presents an open-source toolkit for building recommender systems that supports multiple data processing backends and hardware architectures. https://github.com/sb-ai-lab/RePlay https://arxiv.org/abs/2409.07272 https://github.com/sb-ai-lab/RePlay https://arxiv.org/abs/2409.07272"
X Link 2024-09-12T02:39Z [----] followers, [----] engagements
"Negative Sampling in Recommendation: A Survey and Future Directions Comprehensively reviews negative sampling strategies in recommender systems categorizing existing methods discussing challenges and future directions. https://github.com/hulkima/NS4RS https://arxiv.org/abs/2409.07237 https://github.com/hulkima/NS4RS https://arxiv.org/abs/2409.07237"
X Link 2024-09-12T02:49Z [----] followers, [---] engagements
"Enhancing Q&A Text Retrieval with Ranking Models: Benchmarking fine-tuning and deploying Rerankers for RAG NVIDIA benchmarks ranking models for text retrieval in QA tasks introducing a new sota model NV-RerankQA-Mistral-4B-v3 https://build.nvidia.com/explore/retrieval#nv-rerankqa-mistral-4b-v3 https://arxiv.org/abs/2409.07691 https://build.nvidia.com/explore/retrieval#nv-rerankqa-mistral-4b-v3 https://arxiv.org/abs/2409.07691"
X Link 2024-09-13T03:43Z [----] followers, [----] engagements
"Agents in Software Engineering: Survey Landscape and Vision Presents a comprehensive survey on LLM-based agents in software engineering introducing a framework with perception memory and action modules. https://github.com/DeepSoftwareAnalytics/Awesome-Agent4SE https://arxiv.org/abs/2409.09030 https://github.com/DeepSoftwareAnalytics/Awesome-Agent4SE https://arxiv.org/abs/2409.09030"
X Link 2024-09-16T04:25Z [----] followers, [----] engagements
"RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval Accelerates attention computation in LLMs by using a vector search-based approach to retrieve key-value pairs from CPU memory. https://github.com/mit-han-lab/quest https://arxiv.org/abs/2409.10516 https://github.com/mit-han-lab/quest https://arxiv.org/abs/2409.10516"
X Link 2024-09-17T04:04Z [----] followers, [----] engagements
"beeFormer: Bridging the Gap Between Semantic and Interaction Similarity in Recommender Systems Introduces a framework for training sentence Transformer models on interaction data with text-side information. https://github.com/recombee/beeformer https://arxiv.org/abs/2409.10309 https://github.com/recombee/beeformer https://arxiv.org/abs/2409.10309"
X Link 2024-09-17T04:11Z [----] followers, [---] engagements
"Trustworthiness in Retrieval-Augmented Generation Systems: A Survey Proposes a unified framework to evaluate the trustworthiness of RAG systems across six dimensions offering benchmarks and insights. https://github.com/smallporridge/TrustworthyRAG https://arxiv.org/abs/2409.10102 https://github.com/smallporridge/TrustworthyRAG https://arxiv.org/abs/2409.10102"
X Link 2024-09-17T04:20Z [----] followers, [----] engagements
"HyPA-RAG: A Hybrid Parameter Adaptive Retrieval-Augmented Generation System for AI Legal and Policy Applications Presents a hybrid RAG system with adaptive parameter tuning and combined retrieval methods improving accuracy and fidelity. https://arxiv.org/abs/2409.09046 https://arxiv.org/abs/2409.09046"
X Link 2024-09-17T04:44Z [----] followers, [---] engagements
"Promptriever: Instruction-Trained Retrievers Can Be Prompted Like Language Models Proposes a retrieval model that follows natural language instructions enabling more flexible and user-friendly search experiences. https://github.com/orionw/promptriever https://arxiv.org/abs/2409.11136 https://github.com/orionw/promptriever https://arxiv.org/abs/2409.11136"
X Link 2024-09-18T04:53Z [----] followers, [----] engagements
"Measuring and Enhancing Trustworthiness of LLMs in RAG through Grounded Attributions and Learning to Refuse Introduces a metric for evaluating LLM trustworthiness in RAG systems and a framework to improve grounded responses. https://github.com/declare-lab/trust-align https://arxiv.org/abs/2409.11242 https://github.com/declare-lab/trust-align https://arxiv.org/abs/2409.11242"
X Link 2024-09-18T05:03Z [----] followers, [----] engagements
"Retrieve Annotate Evaluate Repeat: Leveraging Multimodal LLMs for Large-Scale Product Retrieval Evaluation Zalando presents a framework using Multimodal LLMs to efficiently evaluate large-scale e-commerce product retrieval systems. https://arxiv.org/abs/2409.11860 https://arxiv.org/abs/2409.11860"
X Link 2024-09-19T02:46Z [----] followers, [----] engagements
"Fact Fetch and Reason: A Unified Evaluation of Retrieval-Augmented Generation Introduces an evaluation dataset designed to test RAG systems' capabilities in factual accuracy retrieval and reasoning. https://huggingface.co/datasets/google/frames-benchmark https://arxiv.org/abs/2409.12941 https://huggingface.co/datasets/google/frames-benchmark https://arxiv.org/abs/2409.12941"
X Link 2024-09-20T03:14Z [----] followers, [----] engagements
"HLLM: Enhancing Sequential Recommendations via Hierarchical Large Language Models for Item and User Modeling ByteDance introduces a Hierarchical LLM architecture for sequential recommendation systems. https://github.com/bytedance/HLLM https://arxiv.org/abs/2409.12740 https://github.com/bytedance/HLLM https://arxiv.org/abs/2409.12740"
X Link 2024-09-20T03:21Z [----] followers, [----] engagements
"Retrieval-Augmented Test Generation: How Far Are We Investigates the use of RAG for unit test generation comparing different knowledge sources like API documentation GitHub issues and StackOverflow Q&As. https://anonymous.4open.science/r/api_guided_testgen-FB88/README.md https://arxiv.org/abs/2409.12682 https://anonymous.4open.science/r/api_guided_testgen-FB88/README.md https://arxiv.org/abs/2409.12682"
X Link 2024-09-20T03:28Z [----] followers, [---] engagements
"RAD-Bench: Evaluating Large Language Models Capabilities in Retrieval Augmented Dialogues Introduces a benchmark designed to evaluate LLMs in multi-turn dialogues using RAG focusing on retrieval synthesis and reasoning. https://github.com/mtkresearch/RAD-Bench https://arxiv.org/abs/2409.12558 https://github.com/mtkresearch/RAD-Bench https://arxiv.org/abs/2409.12558"
X Link 2024-09-20T03:32Z [----] followers, [---] engagements
"Data Augmentation for Sequential Recommendation: A Survey Comprehensively reviews data augmentation methods for sequential recommendation systems. https://github.com/KingGugu/DA-CL-4Rec https://arxiv.org/abs/2409.13545 https://github.com/KingGugu/DA-CL-4Rec https://arxiv.org/abs/2409.13545"
X Link 2024-09-23T04:47Z [----] followers, [----] engagements
"Contextual Compression in Retrieval-Augmented Generation for Large Language Models: A Survey Examines contextual compression techniques for LLMs focusing on their use in RAG systems and also presents a new taxonomy. https://github.com/SrGrace/Contextual-Compression https://arxiv.org/abs/2409.13385 https://github.com/SrGrace/Contextual-Compression https://arxiv.org/abs/2409.13385"
X Link 2024-09-23T04:50Z [----] followers, [----] engagements
"Retrieval Augmented Generation (RAG) and Beyond: A Comprehensive Survey on How to Make your LLMs use External Data More Wisely Microsoft categorizes data-augmented LLM queries and proposes strategies to tackle challenges in specialized domains. https://arxiv.org/abs/2409.14924 https://arxiv.org/abs/2409.14924"
X Link 2024-09-24T04:34Z [----] followers, 14.2K engagements
"Exploring Hint Generation Approaches in Open-Domain Question Answering Presents a context preparation method for QA systems that uses automatic hint generation instead of traditional retrieval or generation approaches. https://github.com/DataScienceUIBK/HintQA https://arxiv.org/abs/2409.16096 https://github.com/DataScienceUIBK/HintQA https://arxiv.org/abs/2409.16096"
X Link 2024-09-25T02:48Z [----] followers, [---] engagements
"Making Text Embedders Few-Shot Learners Introduces a text embedding model that leverages LLMs' in-context learning capabilities to generate high-quality adaptable embeddings. https://github.com/FlagOpen/FlagEmbedding https://arxiv.org/abs/2409.15700 https://github.com/FlagOpen/FlagEmbedding https://arxiv.org/abs/2409.15700"
X Link 2024-09-25T02:55Z [----] followers, [---] engagements
"Enhancing Performance and Scalability of Large-Scale Recommendation Systems with Jagged Flash Attention Meta introduces Jagged Interaction Kernels and Jagged Flash Attention optimizing recommender systems for variable-length categorical features. https://arxiv.org/abs/2409.15373 https://arxiv.org/abs/2409.15373"
X Link 2024-09-25T03:10Z [----] followers, [---] engagements
"Block-Attention for Low-Latency RAG Significantly reduces inference latency in RAG by pre-computing and caching key-value states for independent blocks of input achieving comparable accuracy to traditional models. https://github.com/TemporaryLoRA/Block-Attention https://arxiv.org/abs/2409.15355 https://github.com/TemporaryLoRA/Block-Attention https://arxiv.org/abs/2409.15355"
X Link 2024-09-25T03:13Z [----] followers, [---] engagements
"FusionANNS: An Efficient CPU/GPU Cooperative Processing Architecture for Billion-scale Approximate Nearest Neighbor Search Presents a high-performance ANNS system for billion-scale datasets that uses CPU/GPU collaboration and SSD storage. https://arxiv.org/abs/2409.16576c https://arxiv.org/abs/2409.16576c"
X Link 2024-09-26T02:46Z [----] followers, [----] engagements
"Disentangling Questions from Query Generation for Task-Adaptive Retrieval Presents a query generation system that adapts to diverse search intents using meta-prompts and retriever feedback. https://github.com/lilys012/metaprompt-QG https://arxiv.org/abs/2409.16570 https://github.com/lilys012/metaprompt-QG https://arxiv.org/abs/2409.16570"
X Link 2024-09-26T02:49Z [----] followers, [---] engagements
"Mixed-Precision Embeddings for Large-Scale Recommendation Models Presents a method for compressing embedding tables in recommender systems assigning varying precisions to feature groups based on importance. https://github.com/Leopold1423/mpe https://arxiv.org/abs/2409.20305 https://github.com/Leopold1423/mpe https://arxiv.org/abs/2409.20305"
X Link 2024-10-02T01:26Z [----] followers, [----] engagements
"Crafting Personalized Agents through Retrieval-Augmented Generation on Editable Memory Graphs Huawei proposes creating personalized AI agents that leverage users' smartphone memories to enhance LLM capabilities. https://arxiv.org/abs/2409.19401 https://arxiv.org/abs/2409.19401"
X Link 2024-10-02T01:52Z [----] followers, [---] engagements
"Winning Solution For Meta KDD Cup' [--] Describes the winning solutions for the KDD Cup [--] RAG challenge detailing a web retrieval framework with tuned LLMs and a regularized API approach achieving first place in all [--] tasks. https://gitlab.aicrowd.com/jiazunchen/kdd2024cup-crag-db3 https://arxiv.org/abs/2410.00005 https://gitlab.aicrowd.com/jiazunchen/kdd2024cup-crag-db3 https://arxiv.org/abs/2410.00005"
X Link 2024-10-02T02:41Z [----] followers, [---] engagements
"Retro-li: Small-Scale Retrieval Augmented Generation Supporting Noisy Similarity Searches and Domain Shift Generalization Improves RAG systems using a small-scale database with enhanced semantic search and noise regularization https://github.com/IBM/Retrieval-Enhanced-Transformer-Little https://arxiv.org/abs/2410.00004 https://github.com/IBM/Retrieval-Enhanced-Transformer-Little https://arxiv.org/abs/2410.00004"
X Link 2024-10-02T02:45Z [----] followers, [---] engagements
"Open-RAG: Enhanced Retrieval-Augmented Reasoning with Open-Source Large Language Models Enhances reasoning in RAG using open-source LLMs by transforming them into sparse mixture of experts models with adaptive retrieval. https://openragmoe.github.io/ https://arxiv.org/abs/2410.01782 https://openragmoe.github.io/ https://arxiv.org/abs/2410.01782"
X Link 2024-10-03T03:18Z [----] followers, [---] engagements
"Elaborative Subtopic Query Reformulation for Broad and Indirect Queries in Travel Destination Recommendation Introduces an LLM-based query reformulation method that improves travel recommender systems. https://github.com/YifanLiu2/ROEGEN-RecSys-24-EQR https://arxiv.org/abs/2410.01598 https://github.com/YifanLiu2/ROEGEN-RecSys-24-EQR https://arxiv.org/abs/2410.01598"
X Link 2024-10-03T03:33Z [----] followers, [----] engagements
"Contextual Document Embeddings Improves neural retrieval by incorporating document context through a new contrastive learning objective and a context-aware architecture. https://arxiv.org/abs/2410.02525 https://arxiv.org/abs/2410.02525"
X Link 2024-10-04T04:32Z [----] followers, [----] engagements
"Dreaming User Multimodal Representation for Micro-Video Recommendation Kuaishou models user interests in a unified multimodal space leveraging historical interactions and addressing cold-start scenarios. https://arxiv.org/abs/2410.03538 https://arxiv.org/abs/2410.03538"
X Link 2024-10-07T02:59Z [----] followers, [---] engagements
"Inductive Generative Recommendation via Retrieval-based Speculation Enables generative recommendation models to recommend new unseen items by combining an inductive drafter with a GR model verifier. https://github.com/Jamesding000/SpecGR https://arxiv.org/abs/2410.02939 https://github.com/Jamesding000/SpecGR https://arxiv.org/abs/2410.02939"
X Link 2024-10-07T03:21Z [----] followers, [----] engagements
"Deciphering the Interplay of Parametric and Non-parametric Memory in Retrieval-augmented Language Models Examines how the Atlas RAG model balances parametric and non-parametric knowledge finding it favors retrieved context. https://github.com/m3hrdadfi/rag-memory-interplay https://arxiv.org/abs/2410.05162 https://github.com/m3hrdadfi/rag-memory-interplay https://arxiv.org/abs/2410.05162"
X Link 2024-10-08T05:46Z [----] followers, [---] engagements
"TableRAG: Million-Token Table Understanding with Language Models Enables efficient large-scale table understanding for language models using smart retrieval techniques to overcome context length limitations while reducing token consumption. https://arxiv.org/abs/2410.04739 https://arxiv.org/abs/2410.04739"
X Link 2024-10-08T05:55Z [----] followers, [----] engagements
"Enhancing Playback Performance in Video Recommender Systems with an On-Device Gating and Ranking Framework Kuaishou introduces an on-device Gating and Ranking Framework to address choppy video playback in recommender systems. https://arxiv.org/abs/2410.05863 https://arxiv.org/abs/2410.05863"
X Link 2024-10-10T07:01Z [----] followers, [---] engagements
"Meta Learning to Rank for Sparsely Supervised Queries Presents a meta-learning to rank framework that effectively addresses sparsely supervised queries enabling quick adaptation to new queries through its ability to generate query-specific rankers. https://dl.acm.org/doi/10.1145/3698876 https://dl.acm.org/doi/10.1145/3698876"
X Link 2024-10-11T04:01Z [----] followers, [----] engagements
"Retriever-and-Memory: Towards Adaptive Note-Enhanced Retrieval-Augmented Generation Uses iterative information collection and adaptive memory review to improve knowledge integration and answer quality. https://github.com/thunlp/Adaptive-Note https://arxiv.org/abs/2410.08821 https://github.com/thunlp/Adaptive-Note https://arxiv.org/abs/2410.08821"
X Link 2024-10-14T02:23Z [----] followers, [----] engagements
"VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents Introduces a VLM-based RAG system that processes multi-modal documents as images achieving up to 39% performance improvement in retrieval tasks. https://github.com/openbmb/visrag https://arxiv.org/abs/2410.10594 https://github.com/openbmb/visrag https://arxiv.org/abs/2410.10594"
X Link 2024-10-15T04:35Z [----] followers, [---] engagements
"Learning to Rank for Multiple Retrieval-Augmented Models through Iterative Utility Maximization Presents an iterative framework for optimizing a unified search engine serving multiple RAG agents. https://github.com/alirezasalemi7/uRAG https://arxiv.org/abs/2410.09942 https://github.com/alirezasalemi7/uRAG https://arxiv.org/abs/2410.09942"
X Link 2024-10-15T04:55Z [----] followers, [---] engagements
"Agentic Information Retrieval Proposes a paradigm using LLM agents to expand and transform traditional information retrieval offering a unified flexible approach to complex information tasks. https://arxiv.org/abs/2410.09713 https://arxiv.org/abs/2410.09713"
X Link 2024-10-15T05:05Z [----] followers, [----] engagements
"Synthetic Knowledge Ingestion: Towards Knowledge Refinement and Injection for Enhancing Large Language Models Intuit enhances LLMs' knowledge capabilities through fine-grained synthesis interleaved generation and assemble augmentation. https://arxiv.org/abs/2410.09629 https://arxiv.org/abs/2410.09629"
X Link 2024-10-15T05:08Z [----] followers, [---] engagements
"Toward General Instruction-Following Alignment for Retrieval-Augmented Generation Introduces an automated pipeline for improving instruction-following in RAG systems alongside FollowRAG a new benchmark for evaluation. https://followrag.github.io/ https://arxiv.org/abs/2410.09584 https://followrag.github.io/ https://arxiv.org/abs/2410.09584"
X Link 2024-10-15T05:12Z [----] followers, [---] engagements
"Your Mixture-of-Experts LLM Is Secretly an Embedding Model For Free Combines routing weights and hidden states from MoE LLMs to create superior embeddings without additional training. https://github.com/tianyi-lab/MoE-Embedding https://arxiv.org/abs/2410.10814 https://github.com/tianyi-lab/MoE-Embedding https://arxiv.org/abs/2410.10814"
X Link 2024-10-15T05:18Z [----] followers, 12.9K engagements
"Sequential LLM Framework for Fashion Recommendation Presents a sequential fashion recommendation system using an LLM employing specialized prompts efficient fine-tuning and a mix-up-based retrieval technique. https://arxiv.org/abs/2410.11327 https://arxiv.org/abs/2410.11327"
X Link 2024-10-16T05:16Z [----] followers, [---] engagements
"FRAG: Toward Federated Vector Database Management for Collaborative and Secure Retrieval-Augmented Generation Enables secure efficient collaboration by allowing mutually distrusting parties to perform encrypted ANN searche across distributed databases https://arxiv.org/abs/2410.13272 https://arxiv.org/abs/2410.13272"
X Link 2024-10-18T06:22Z [----] followers, [---] engagements
"Starbucks: Improved Training for 2D Matryoshka Embeddings Improves 2D Matryoshka by using targeted layer-dimension pairs achieving effectiveness comparable to separately trained models while maintaining adaptability. https://github.com/ielab/Starbucks https://arxiv.org/abs/2410.13230 https://github.com/ielab/Starbucks https://arxiv.org/abs/2410.13230"
X Link 2024-10-18T06:32Z [----] followers, [----] engagements
"Optimizing and Evaluating Enterprise Retrieval-Augmented Generation (RAG): A Content Design Perspective IBM details practical experiences in building enterprise-scale RAG solutions for software documentation. https://github.com/spackows/ICAAI-2024_RAG-CD https://arxiv.org/abs/2410.12812 https://github.com/spackows/ICAAI-2024_RAG-CD https://arxiv.org/abs/2410.12812"
X Link 2024-10-18T06:45Z [----] followers, [---] engagements
"Simplify to the Limit Embedding-less Graph Collaborative Filtering for Recommender Systems Improves on GCN and GCL approaches by focusing on user-item similarity eliminating embeddings and using contrastive objectives. https://github.com/BlueGhostYi/ID-GRec https://dl.acm.org/doi/10.1145/3701230 https://github.com/BlueGhostYi/ID-GRec https://dl.acm.org/doi/10.1145/3701230"
X Link 2024-10-21T02:48Z [----] followers, [----] engagements
"Centrality-aware Product Retrieval and Ranking Introduces an approach to improve e-commerce product search by using dual-loss optimization to handle semantically relevant but intent-mismatched results. https://arxiv.org/abs/2410.15930 https://arxiv.org/abs/2410.15930"
X Link 2024-10-23T04:08Z [----] followers, [----] engagements
"HyQE: Ranking Contexts with Hypothetical Query Embeddings Intuit presents a scalable context ranking framework that uses LLMs to generate hypothetical queries from retrieved contexts and ranks them. https://github.com/zwc662/hyqe https://arxiv.org/abs/2410.15262 https://github.com/zwc662/hyqe https://arxiv.org/abs/2410.15262"
X Link 2024-10-23T04:31Z [----] followers, [---] engagements
"Class-RAG: Content Moderation with Retrieval Augmented Generation Meta GenAI Team introduces a flexible content moderation system that uses RAG to improve classification accuracy and adaptability. https://arxiv.org/abs/2410.14881 https://arxiv.org/abs/2410.14881"
X Link 2024-10-23T04:39Z [----] followers, [---] engagements
"Bridging Search and Recommendation in Generative Retrieval: Does One Task Help the Other Spotify demonstrates how unifying search and recommendation tasks into a single LLM-based generative model can outperform specialized models. https://arxiv.org/abs/2410.16823 https://arxiv.org/abs/2410.16823"
X Link 2024-10-23T05:02Z [----] followers, [----] engagements
"SouLLMate: An Application Enhancing Diverse Mental Health Support with Adaptive LLMs Prompt Engineering and RAG Techniques Presents an LLM-powered mental health support system achieving 80% accuracy in clinical assessments. https://github.com/QM378/SouLLMate https://arxiv.org/abs/2410.16322 https://github.com/QM378/SouLLMate https://arxiv.org/abs/2410.16322"
X Link 2024-10-23T05:13Z [----] followers, [---] engagements
"DeCoRe: Decoding by Contrasting Retrieval Heads to Mitigate Hallucinations Reduces hallucinations in LLMs by contrasting outputs between normal and deliberately impaired versions of the model (via masked retrieval heads). https://github.com/aryopg/DeCoRe https://arxiv.org/abs/2410.18860 https://github.com/aryopg/DeCoRe https://arxiv.org/abs/2410.18860"
X Link 2024-10-25T04:22Z [----] followers, [----] engagements
"Little Giants: Synthesizing High-Quality Embedding Data at Scale Trains small (8B) language models to generate high-quality synthetic data for text embeddings matching the performance of GPT-4-based approaches with less than 1/10 of the API calls. https://arxiv.org/abs/2410.18634 https://arxiv.org/abs/2410.18634"
X Link 2024-10-25T04:25Z [----] followers, [----] engagements
"Quam: Adaptive Retrieval through Query Affinity Modelling Presents an improved adaptive retrieval approach that uses query affinity modeling to achieve better recall over standard re-ranking methods. https://github.com/Mandeep-Rathee/quam https://arxiv.org/abs/2410.20286 https://github.com/Mandeep-Rathee/quam https://arxiv.org/abs/2410.20286"
X Link 2024-10-29T05:03Z [----] followers, [---] engagements
"Enhancing CTR prediction in Recommendation Domain with Search Query Representation Huawei introduces a framework that combines search query embeddings and collaborative filtering to boost CTR prediction by learning from users' search behavior patterns. https://arxiv.org/abs/2410.21487 https://arxiv.org/abs/2410.21487"
X Link 2024-10-30T03:25Z [----] followers, [---] engagements
"Semantic Search Evaluation LinkedIn presents a method for evaluating content search systems by measuring semantic relevance between queries and results using LLMs introducing an "on-topic rate" metric for quality assessment. https://arxiv.org/abs/2410.21549 https://arxiv.org/abs/2410.21549"
X Link 2024-10-30T03:25Z [----] followers, [----] engagements
"LLM-KT: A Versatile Framework for Knowledge Transfer from Large Language Models to Collaborative Filtering Enhances collaborative filtering models by injecting LLM-generated features into intermediate layers. https://github.com/a250/LLMRecSys_with_KnowledgeDistilation/tree/distil_framework https://arxiv.org/abs/2411.00556 https://github.com/a250/LLMRecSys_with_KnowledgeDistilation/tree/distil_framework https://arxiv.org/abs/2411.00556"
X Link 2024-11-04T04:11Z [----] followers, [----] engagements
"Beyond Utility: Evaluating LLM as Recommender Proposes an evaluation framework for LLM-based recommender systems covering utility history length sensitivity candidate position bias generation quality and hallucinations. https://github.com/JiangDeccc/EvaLLMasRecommender https://arxiv.org/abs/2411.00331 https://github.com/JiangDeccc/EvaLLMasRecommender https://arxiv.org/abs/2411.00331"
X Link 2024-11-04T04:17Z [----] followers, [---] engagements
"Rationale-Guided Retrieval Augmented Generation for Medical Question Answering Proposes a retrieval framework that enhances medical QA by filtering retrieved documents based on rationale perplexity. https://github.com/dmis-lab/RAG2 https://arxiv.org/abs/2411.00300 https://github.com/dmis-lab/RAG2 https://arxiv.org/abs/2411.00300"
X Link 2024-11-04T04:24Z [----] followers, [----] engagements
"PSL: Rethinking and Improving Softmax Loss from Pairwise Perspective for Recommendation Introduces Pairwise Softmax Loss that provides tighter bounds for ranking metrics and better robustness against noise and dist shifts. https://github.com/Tiny-Snow/IR-Benchmark https://arxiv.org/abs/2411.00163 https://github.com/Tiny-Snow/IR-Benchmark https://arxiv.org/abs/2411.00163"
X Link 2024-11-04T04:25Z [----] followers, [----] engagements
"JudgeRank: Leveraging Large Language Models for Reasoning-Intensive Reranking Introduces an agentic reranker that improves document retrieval by analyzing queries and documents through explicit reasoning steps. https://arxiv.org/abs/2411.00142 https://arxiv.org/abs/2411.00142"
X Link 2024-11-04T04:29Z [----] followers, [----] engagements
"LLM4PR: Improving Post-Ranking in Search Engine with Large Language Models Kuaishou introduces the first LLM-based framework for post-ranking optimization in search engines leveraging a Query-Instructed Adapter and feature adaptation steps. https://arxiv.org/abs/2411.01178 https://arxiv.org/abs/2411.01178"
X Link 2024-11-05T05:24Z [----] followers, [---] engagements
"MM-EMBED: Universal Multimodal Retrieval with Multimodal LLMs NVIDIA presents a versatile multimodal retriever that achieves SOTA on both multimodal and text retrieval tasks while maintaining strong text-to-text capabilities. https://huggingface.co/nvidia/MM-Embed https://arxiv.org/abs/2411.02571 https://huggingface.co/nvidia/MM-Embed https://arxiv.org/abs/2411.02571"
X Link 2024-11-06T04:28Z [----] followers, [----] engagements
"Benchmarking Multimodal Retrieval Augmented Generation with Dynamic VQA Dataset and Self-Adaptive Planning Agent Alibaba introduces a self-adaptive planning agent for multimodal retrieval along with a new dataset. https://github.com/Alibaba-NLP/OmniSearch https://arxiv.org/abs/2411.02937 https://github.com/Alibaba-NLP/OmniSearch https://arxiv.org/abs/2411.02937"
X Link 2024-11-06T04:29Z [----] followers, [---] engagements
"RAG-QA Arena: Evaluating Domain Robustness for Long-Form Retrieval-Augmented Question Answering Aamazon presents a new dataset with human-written long-form answers across [--] domains and proposes an evaluation framework. https://github.com/awslabs/rag-qa-arena https://www.amazon.science/publications/rag-qa-arena-evaluating-domain-robustness-for-long-form-retrieval-augmented-question-answering https://github.com/awslabs/rag-qa-arena https://www.amazon.science/publications/rag-qa-arena-evaluating-domain-robustness-for-long-form-retrieval-augmented-question-answering"
X Link 2024-11-06T04:34Z [----] followers, [----] engagements
"HtmlRAG: HTML is Better Than Plain Text for Modeling Retrieved Knowledge in RAG Systems Leverages HTML structures instead of plain text for RAG with effective HTML cleaning and pruning strategies to maintain efficiency. https://github.com/plageon/HtmlRAG https://arxiv.org/abs/2411.02959 https://github.com/plageon/HtmlRAG https://arxiv.org/abs/2411.02959"
X Link 2024-11-06T04:35Z [----] followers, [---] engagements
"RAGulator: Lightweight Out-of-Context Detectors for Grounded Text Generation Introduces a lightweight approach to detect out-of-context outputs in RAG systems using discriminative models trained on minimal resources. https://arxiv.org/abs/2411.03920 https://arxiv.org/abs/2411.03920"
X Link 2024-11-07T04:06Z [----] followers, [----] engagements
"Long Context RAG Performance of Large Language Models Databricks analyzes [--] LLMs and reveals that only recent state-of-the-art models maintain consistent RAG accuracy above 64k tokens with most models' performance declining at longer contexts. https://arxiv.org/abs/2411.03538 https://arxiv.org/abs/2411.03538"
X Link 2024-11-07T04:08Z [----] followers, 12.5K engagements
"Lightning IR: Straightforward Fine-tuning and Inference of Transformer-based Language Models for Information Retrieval Introduces a PyTorch Lightning-based framework for fine-tuning and inference of transformer models in IR. https://github.com/webis-de/lightning-ir https://arxiv.org/abs/2411.04677 https://github.com/webis-de/lightning-ir https://arxiv.org/abs/2411.04677"
X Link 2024-11-08T04:27Z [----] followers, [----] engagements
"Best Practices for Distilling Large Language Models into BERT for Web Search Ranking Tencent introduces a distillation framework that effectively transfers LLM ranking capabilities to BERT models while dramatically reducing inference costs. https://arxiv.org/abs/2411.04539 https://arxiv.org/abs/2411.04539"
X Link 2024-11-08T04:28Z [----] followers, 10.6K engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing