[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.] #  @HuggingPapers DailyPapers DailyPapers posts on X about ai, bytedance, native, unified the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXXX engagements in the last XX hours. ### Engagements: XXXXXX [#](/creator/twitter::1906617820122595328/interactions)  - X Week XXXXXXX -XX% - X Month XXXXXXX -XXXX% - X Months XXXXXXXXX +874% ### Mentions: XX [#](/creator/twitter::1906617820122595328/posts_active)  - X Week XX -XX% - X Month XXX -XX% - X Months XXX +1,330% ### Followers: XXXXXX [#](/creator/twitter::1906617820122595328/followers)  - X Week XXXXXX +3.50% - X Month XXXXXX +19% - X Months XXXXXX +271% ### CreatorRank: XXXXXXX [#](/creator/twitter::1906617820122595328/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) [stocks](/list/stocks) [social networks](/list/social-networks) [finance](/list/finance) [currencies](/list/currencies) **Social topic influence** [ai](/topic/ai), [bytedance](/topic/bytedance) #16, [native](/topic/native) #715, [unified](/topic/unified) #30, [llm](/topic/llm) #161, [math](/topic/math), [alibaba](/topic/alibaba), [generative](/topic/generative) #223, [meituan](/topic/meituan) #20, [tencent](/topic/tencent) #238 **Top accounts mentioned or mentioned by** [@huggingface](/creator/undefined) [@alibabacloud](/creator/undefined) [@10](/creator/undefined) [@32](/creator/undefined) [@ostrisai](/creator/undefined) [@bytedanceai](/creator/undefined) [@ucscai](/creator/undefined) [@gameofthrones](/creator/undefined) [@googleais](/creator/undefined) [@internlm](/creator/undefined) [@deepseekai](/creator/undefined) [@alibabaqwen](/creator/undefined) [@nvidia](/creator/undefined) [@brandgrowthos](/creator/undefined) [@codewithimanshu](/creator/undefined) [@viumobile](/creator/undefined) [@kevinqhlin](/creator/undefined) [@latentspacer](/creator/undefined) [@zettaidao](/creator/undefined) **Top assets mentioned** [Alibaba Group (BABA)](/topic/alibaba-group) [Microsoft Corp. (MSFT)](/topic/microsoft) ### Top Social Posts Top posts by engagements in the last XX hours "Alibaba's Qwen3-VL unveils a new era for multimodal AI It's the most capable vision-language model yet featuring native 256K context for text and video enhanced text understanding and advanced reasoning across diverse visual tasks" [X Link](https://x.com/HuggingPapers/status/1996433668479520947) 2025-12-04T04:17Z 10.3K followers, 17.1K engagements "Tencent researchers unveil Deep Research: A Systematic Survey This comprehensive survey maps the evolving field of Deep Research detailing how LLMs combine with external tools (like search engines) to act as powerful verifiable research agents. It covers X key components: query planning info acquisition memory management and answer generation" [X Link](https://x.com/HuggingPapers/status/1997098804269490479) 2025-12-06T00:20Z 10.3K followers, 4302 engagements "Liquid AI presents LFM2: Liquid Foundation Models for efficient on-device AI. This family of models (350M-8.3B) offers up to 2x faster CPU performance for prefill & decode achieving strong benchmark results ideal for memory-efficient edge apps" [X Link](https://x.com/HuggingPapers/status/1997639690871091606) 2025-12-07T12:09Z 10.3K followers, 2689 engagements "Zhipu AI just released GLM-4.6V on Hugging Face This new multimodal model achieves SOTA visual understanding features native function calling for agents and handles 128k context for documents. Perception to action" [X Link](https://x.com/HuggingPapers/status/1998373902595301589) 2025-12-09T12:47Z 10.3K followers, 16.2K engagements "Unified Video Editing with Temporal Reasoner (VideoCoF) This novel Chain-of-Frames approach enables precise mask-free video editing and 4x length extrapolation achieving SOTA performance with just 50k training pairs" [X Link](https://x.com/HuggingPapers/status/1998425189500305845) 2025-12-09T16:11Z 10.3K followers, 4274 engagements "Qwen just launched Qwen3-Next-80B-A3B-Thinking on Hugging Face This new model combines Hybrid Attention & high-sparsity MoE for efficient ultra-long context (1M tokens) and complex reasoning. It outperforms Gemini-2.5-Flash-Thinking" [X Link](https://x.com/HuggingPapers/status/1998645065598775547) 2025-12-10T06:44Z 10.3K followers, 20.2K engagements "Alibaba Group unveils ReForm: Reflective Autoformalization for math It translates natural language math into machine-verifiable Lean4. An iterative "generate validate refine" cycle self-corrects semantic errors boosting performance by XXXX% over baselines" [X Link](https://x.com/HuggingPapers/status/1984053291882570206) 2025-10-31T00:22Z 10.3K followers, 1124 engagements "ByteDance just released Sa2VA on Hugging Face. It's the first unified model for dense grounded understanding of images and videos combining SAM-2 with LLaVA for advanced visual perception" [X Link](https://x.com/HuggingPapers/status/1993984244524453968) 2025-11-27T10:04Z 10.3K followers, 4787 engagements "ByteDance introduces Adv-GRPO This new RL framework for image generation uses adversarial rewards & visual foundation models (like DINO) to combat "reward hacking." It achieves superior image quality & aesthetics by taking the image itself as a dense visual reward" [X Link](https://x.com/HuggingPapers/status/1994741274851905936) 2025-11-29T12:12Z 10.3K followers, 9880 engagements "Alibaba Group unveils Z-Image: an efficient 6B-parameter foundation model for image generation. This new model challenges the "scale-at-all-costs" paradigm offering exceptional photorealistic quality and bilingual text rendering all while being efficient enough for consumer GPUs" [X Link](https://x.com/HuggingPapers/status/1995349763047428237) 2025-12-01T04:30Z 10.3K followers, 2528 engagements "Experience Z-Image-Turbo: sub-second inference (8 steps) on 16GB VRAM. Get the paper model & demo on @HuggingFace: 📄 Paper: 💾 Model: ✨ Demo:" [X Link](https://x.com/HuggingPapers/status/1995349772799246490) 2025-12-01T04:30Z 10.3K followers, XXX engagements "REASONEDIT: Towards Reasoning-Enhanced Image Editing Models just dropped This new framework leverages MLLM thinking & reflection to interpret abstract instructions and iteratively refine image edits pushing the boundaries of what's possible in generative AI" [X Link](https://x.com/HuggingPapers/status/1995404903435419690) 2025-12-01T08:09Z 10.3K followers, 2197 engagements "Tencent's HunyuanVideo XXX is here A lightweight yet powerful open-source video generation model achieving state-of-the-art visual quality & motion coherence with just 8.3B params enabling efficient inference on consumer GPUs" [X Link](https://x.com/HuggingPapers/status/1995163000861319172) 2025-11-30T16:08Z 10.3K followers, 18.1K engagements "Tencent AI Lab unveils R-Few: Self-Evolving LLMs with minimal human guidance Introducing a Challenger-Solver framework for stable self-improvement overcoming issues like concept drift and diversity collapse for language models" [X Link](https://x.com/HuggingPapers/status/1996250609453060207) 2025-12-03T16:10Z 10.3K followers, 2591 engagements "Meituan introduces OneThinker an all-in-one visual reasoning model This generalist MLLM unifies image and video understanding across XX diverse tasks like Q&A grounding tracking and segmentation. It achieves strong performance using EMA-GRPO for multi-task RL" [X Link](https://x.com/HuggingPapers/status/1996673525495578921) 2025-12-04T20:10Z 10.3K followers, 9042 engagements "Explore the new OneThinker model from Meituan unifying image and video reasoning Read the paper: Find models & data:" [X Link](https://x.com/HuggingPapers/status/1996673535276630201) 2025-12-04T20:10Z 10.3K followers, XXX engagements "EditThinker: image editors can now "think" iteratively This new framework by Meituan unlocks deliberative editing for any image model improving instruction-following through a Critique-Refine-Repeat cycle" [X Link](https://x.com/HuggingPapers/status/1997882846929928424) 2025-12-08T04:16Z 10.3K followers, 4409 engagements "Huawei just released EMMA: a unified multimodal AI for understanding generation & editing This efficient architecture sets a new SOTA outperforming larger models while handling complex image edits and diverse tasks. It's a leap for multimodal models" [X Link](https://x.com/HuggingPapers/status/1998185640308228149) 2025-12-09T00:19Z 10.3K followers, 15.4K engagements "Explore its capabilities for rich-text generation visual web search and UI replication. Dive into the details and try it out:" [X Link](https://x.com/HuggingPapers/status/1998373911986409576) 2025-12-09T12:47Z 10.3K followers, XXX engagements "Visionary: Your World Model Carrier on the Web This open web-native platform brings real-time Gaussian Splatting (3DGS 4DGS neural avatars) & mesh rendering directly to your browser with WebGPU & ONNX. Say goodbye to heavy installs & hello to "click-to-run" dynamic 3D content" [X Link](https://x.com/HuggingPapers/status/1998727063620972810) 2025-12-10T12:10Z 10.3K followers, 5409 engagements "Meta AI unveils OneStory: Coherent Multi-Shot Video Generation This framework addresses long-range narrative consistency in videos. It uses adaptive memory and next-shot generation with pretrained I2V models to achieve state-of-the-art coherent storytelling" [X Link](https://x.com/HuggingPapers/status/1998847130379718714) 2025-12-10T20:07Z 10.3K followers, 2684 engagements "Sina Weibo AI Lab unveils VibeThinker-1.5B This 1.5B model challenges the scaling consensus achieving top-tier reasoning on math & coding. It surpasses 400x larger models on benchmarks like AIME25 trained at a fraction of the cost ($7800 vs $294K+)" [X Link](https://x.com/HuggingPapers/status/1988459790603821330) 2025-11-12T04:12Z 10.3K followers, 2299 engagements "OPPO AI Agent Team introduces O-Mem an Omni Memory System for LLM Agents It brings personalized long-horizon self-evolving capabilities to conversational AI by actively profiling users and supporting hierarchical retrieval for adaptive responses" [X Link](https://x.com/HuggingPapers/status/1995103128497315944) 2025-11-30T12:10Z 10.3K followers, 14.4K engagements "A new era of physics-aware video generation is here Introducing NewtonRewards a framework that enforces Newton's Laws in video diffusion models. It achieves physically plausible smooth and temporally coherent motions through verifiable rewards outperforming prior methods" [X Link](https://x.com/HuggingPapers/status/1996011529112625227) 2025-12-03T00:20Z 10.3K followers, 11.7K engagements "ByteDance's Nex-N1: Agentic Models Trained via a Unified Ecosystem Introducing Nex-N1 a groundbreaking platform for LLM agents It scales diverse complex interactive environments for effective policy learning outperforming SoTA models on SWE-bench & tau2" [X Link](https://x.com/HuggingPapers/status/1997035945233727796) 2025-12-05T20:10Z 10.3K followers, 5859 engagements "EditThinker learns to critique results refine instructions & repeat until satisfactory simulating human cognitive loops for better results. Paper:" [X Link](https://x.com/HuggingPapers/status/1997882856794915172) 2025-12-08T04:16Z 10.3K followers, XXX engagements "CAPO: New RL for stable generalized reasoning From Xiaomi & Tsinghua an adaptive curriculum uses positive-only advantage for stable foundations then introduces negative signals for discrimination. It boosts math & multimodal GUI reasoning. Plug-and-play with PPO GRPO RLOO & more" [X Link](https://x.com/HuggingPapers/status/1998062312154865842) 2025-12-08T16:09Z 10.3K followers, 3623 engagements "LongVT: Incentivizing "Thinking with Long Videos" This new framework for LMMs uses Multimodal Chain-of-Tool-Thought enabling global-to-local reasoning in long videos via native video cropping to tackle hallucinations and outperform baselines" [X Link](https://x.com/HuggingPapers/status/1995708469559574621) 2025-12-02T04:15Z 10.3K followers, 1438 engagements "ByteDance Seed introduces DAComp: A benchmark for data agents DAComp challenges AI agents across the full data intelligence lifecycle: from complex Data Engineering (multi-stage SQL pipelines) to open-ended Data Analysis. SOTA models struggle XX% success on DE tasks" [X Link](https://x.com/HuggingPapers/status/1996975509742539037) 2025-12-05T16:10Z 10.3K followers, 1454 engagements "Mistral just dropped Mistral Large X on Hugging Face Their new state-of-the-art multimodal Mixture-of-Experts model boasts 41B active params 675B total and a massive 256k context window for frontier performance" [X Link](https://x.com/HuggingPapers/status/1995880411893772362) 2025-12-02T15:39Z 10.3K followers, 8141 engagements "Kuaishou's Kling Team introduces MultiShotMaster This pioneering framework enables highly controllable multi-shot video generation. It tackles narrative complexity with flexible shot arrangement coherent transitions and custom control for subjects and backgrounds" [X Link](https://x.com/HuggingPapers/status/1996190496666845677) 2025-12-03T12:11Z 10.3K followers, 1436 engagements "Kuaishou Technology introduces ViDiC A new benchmark for MLLMs to caption fine-grained similarities & differences between video pairs. ViDiC-1K dataset: 1000 video pairs 4000+ checklist items across X categories (subject motion style & more)" [X Link](https://x.com/HuggingPapers/status/1996491968117866887) 2025-12-04T08:09Z 10.3K followers, 2337 engagements "Microsoft just launched VibeVoice Realtime on Hugging Face A lightweight streaming text-to-speech model generating initial audible speech in 300ms perfect for live data and LLM conversations" [X Link](https://x.com/HuggingPapers/status/1997351370848964755) 2025-12-06T17:04Z 10.3K followers, 218K engagements "Wikontic: Constructing Wikidata-aligned Ontology-Aware Knowledge Graphs This multi-stage pipeline leverages LLMs to build compact consistent knowledge graphs from raw text boosting reasoning without relying on auxiliary data" [X Link](https://x.com/HuggingPapers/status/1997579328918274406) 2025-12-07T08:09Z 10.3K followers, 6128 engagements "Top AI Papers of The Week (December 1-7): - - DeepSeek-V3.2 by @deepseek_ai: Pushing open LLM frontiers with incredible reasoning & agent performance - A Practical Guide to Code Intelligence covering agents and applications from companies like Microsoft ByteDance & Anthropic - LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling for LMMs - Z-Image: An Efficient Image Generation Foundation Model that rivals commercial models - Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length - DAComp: Benchmarking Data Agents across the Full Data" [X Link](https://x.com/HuggingPapers/status/1997672669311823880) 2025-12-07T14:20Z 10.3K followers, 1929 engagements "ByteDance-Seed just released Adversarial Flow Models on Hugging Face These new generative models unify adversarial and flow models achieving a new SOTA FID of XXXX on ImageNet-256px with a single forward pass" [X Link](https://x.com/HuggingPapers/status/1998011263603904649) 2025-12-08T12:46Z 10.3K followers, 12.1K engagements "Alibaba's Tongyi Lab introduces Wan-Move for precise video motion control Generates high-fidelity 5-second 480p videos with control rivaling commercial tools by making original condition features motion-aware" [X Link](https://x.com/HuggingPapers/status/1998607499218374976) 2025-12-10T04:15Z 10.3K followers, 8520 engagements "Alibaba & Tsinghua just dropped Wan-Move A scalable framework bringing fine-grained point-level motion control to video generation. It produces 5-second 480p videos whose quality rivals commercial tools like Kling XXX Pro's Motion Brush" [X Link](https://x.com/HuggingPapers/status/1998675528149045454) 2025-12-10T08:45Z 10.3K followers, 4622 engagements "Visionary democratizes 3DGS research & deployment by offering a lightweight unified platform right in your browser It lowers barriers to reproduce compare & deploy cutting-edge 3D generative models. Paper: Project page:" [X Link](https://x.com/HuggingPapers/status/1998727073221784050) 2025-12-10T12:10Z 10.3K followers, XXX engagements "It achieves state-of-the-art information retention (86% on MINE-1) and creates KGs using 1000 output tokens. Get the full details on Hugging Face:" [X Link](https://x.com/HuggingPapers/status/1997579338531643677) 2025-12-07T08:10Z 10.3K followers, XXX engagements "ByteDance Seed unveils UniUGP for end-to-end autonomous driving. It's a unified framework that synergizes scene understanding future video generation and trajectory planning. Achieves SOTA in complex long-tail scenarios" [X Link](https://x.com/HuggingPapers/status/1998970477839024154) 2025-12-11T04:17Z 10.3K followers, 1674 engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@HuggingPapers DailyPapers
DailyPapers posts on X about ai, bytedance, native, unified the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXXX engagements in the last XX hours.
Engagements: XXXXXX #
- X Week XXXXXXX -XX%
- X Month XXXXXXX -XXXX%
- X Months XXXXXXXXX +874%
Mentions: XX #
- X Week XX -XX%
- X Month XXX -XX%
- X Months XXX +1,330%
Followers: XXXXXX #
- X Week XXXXXX +3.50%
- X Month XXXXXX +19%
- X Months XXXXXX +271%
CreatorRank: XXXXXXX #
Social Influence
Social category influence technology brands stocks social networks finance currencies
Social topic influence ai, bytedance #16, native #715, unified #30, llm #161, math, alibaba, generative #223, meituan #20, tencent #238
Top accounts mentioned or mentioned by @huggingface @alibabacloud @10 @32 @ostrisai @bytedanceai @ucscai @gameofthrones @googleais @internlm @deepseekai @alibabaqwen @nvidia @brandgrowthos @codewithimanshu @viumobile @kevinqhlin @latentspacer @zettaidao
Top assets mentioned Alibaba Group (BABA) Microsoft Corp. (MSFT)
Top Social Posts
Top posts by engagements in the last XX hours
"Alibaba's Qwen3-VL unveils a new era for multimodal AI It's the most capable vision-language model yet featuring native 256K context for text and video enhanced text understanding and advanced reasoning across diverse visual tasks"
X Link 2025-12-04T04:17Z 10.3K followers, 17.1K engagements
"Tencent researchers unveil Deep Research: A Systematic Survey This comprehensive survey maps the evolving field of Deep Research detailing how LLMs combine with external tools (like search engines) to act as powerful verifiable research agents. It covers X key components: query planning info acquisition memory management and answer generation"
X Link 2025-12-06T00:20Z 10.3K followers, 4302 engagements
"Liquid AI presents LFM2: Liquid Foundation Models for efficient on-device AI. This family of models (350M-8.3B) offers up to 2x faster CPU performance for prefill & decode achieving strong benchmark results ideal for memory-efficient edge apps"
X Link 2025-12-07T12:09Z 10.3K followers, 2689 engagements
"Zhipu AI just released GLM-4.6V on Hugging Face This new multimodal model achieves SOTA visual understanding features native function calling for agents and handles 128k context for documents. Perception to action"
X Link 2025-12-09T12:47Z 10.3K followers, 16.2K engagements
"Unified Video Editing with Temporal Reasoner (VideoCoF) This novel Chain-of-Frames approach enables precise mask-free video editing and 4x length extrapolation achieving SOTA performance with just 50k training pairs"
X Link 2025-12-09T16:11Z 10.3K followers, 4274 engagements
"Qwen just launched Qwen3-Next-80B-A3B-Thinking on Hugging Face This new model combines Hybrid Attention & high-sparsity MoE for efficient ultra-long context (1M tokens) and complex reasoning. It outperforms Gemini-2.5-Flash-Thinking"
X Link 2025-12-10T06:44Z 10.3K followers, 20.2K engagements
"Alibaba Group unveils ReForm: Reflective Autoformalization for math It translates natural language math into machine-verifiable Lean4. An iterative "generate validate refine" cycle self-corrects semantic errors boosting performance by XXXX% over baselines"
X Link 2025-10-31T00:22Z 10.3K followers, 1124 engagements
"ByteDance just released Sa2VA on Hugging Face. It's the first unified model for dense grounded understanding of images and videos combining SAM-2 with LLaVA for advanced visual perception"
X Link 2025-11-27T10:04Z 10.3K followers, 4787 engagements
"ByteDance introduces Adv-GRPO This new RL framework for image generation uses adversarial rewards & visual foundation models (like DINO) to combat "reward hacking." It achieves superior image quality & aesthetics by taking the image itself as a dense visual reward"
X Link 2025-11-29T12:12Z 10.3K followers, 9880 engagements
"Alibaba Group unveils Z-Image: an efficient 6B-parameter foundation model for image generation. This new model challenges the "scale-at-all-costs" paradigm offering exceptional photorealistic quality and bilingual text rendering all while being efficient enough for consumer GPUs"
X Link 2025-12-01T04:30Z 10.3K followers, 2528 engagements
"Experience Z-Image-Turbo: sub-second inference (8 steps) on 16GB VRAM. Get the paper model & demo on @HuggingFace: 📄 Paper: 💾 Model: ✨ Demo:"
X Link 2025-12-01T04:30Z 10.3K followers, XXX engagements
"REASONEDIT: Towards Reasoning-Enhanced Image Editing Models just dropped This new framework leverages MLLM thinking & reflection to interpret abstract instructions and iteratively refine image edits pushing the boundaries of what's possible in generative AI"
X Link 2025-12-01T08:09Z 10.3K followers, 2197 engagements
"Tencent's HunyuanVideo XXX is here A lightweight yet powerful open-source video generation model achieving state-of-the-art visual quality & motion coherence with just 8.3B params enabling efficient inference on consumer GPUs"
X Link 2025-11-30T16:08Z 10.3K followers, 18.1K engagements
"Tencent AI Lab unveils R-Few: Self-Evolving LLMs with minimal human guidance Introducing a Challenger-Solver framework for stable self-improvement overcoming issues like concept drift and diversity collapse for language models"
X Link 2025-12-03T16:10Z 10.3K followers, 2591 engagements
"Meituan introduces OneThinker an all-in-one visual reasoning model This generalist MLLM unifies image and video understanding across XX diverse tasks like Q&A grounding tracking and segmentation. It achieves strong performance using EMA-GRPO for multi-task RL"
X Link 2025-12-04T20:10Z 10.3K followers, 9042 engagements
"Explore the new OneThinker model from Meituan unifying image and video reasoning Read the paper: Find models & data:"
X Link 2025-12-04T20:10Z 10.3K followers, XXX engagements
"EditThinker: image editors can now "think" iteratively This new framework by Meituan unlocks deliberative editing for any image model improving instruction-following through a Critique-Refine-Repeat cycle"
X Link 2025-12-08T04:16Z 10.3K followers, 4409 engagements
"Huawei just released EMMA: a unified multimodal AI for understanding generation & editing This efficient architecture sets a new SOTA outperforming larger models while handling complex image edits and diverse tasks. It's a leap for multimodal models"
X Link 2025-12-09T00:19Z 10.3K followers, 15.4K engagements
"Explore its capabilities for rich-text generation visual web search and UI replication. Dive into the details and try it out:"
X Link 2025-12-09T12:47Z 10.3K followers, XXX engagements
"Visionary: Your World Model Carrier on the Web This open web-native platform brings real-time Gaussian Splatting (3DGS 4DGS neural avatars) & mesh rendering directly to your browser with WebGPU & ONNX. Say goodbye to heavy installs & hello to "click-to-run" dynamic 3D content"
X Link 2025-12-10T12:10Z 10.3K followers, 5409 engagements
"Meta AI unveils OneStory: Coherent Multi-Shot Video Generation This framework addresses long-range narrative consistency in videos. It uses adaptive memory and next-shot generation with pretrained I2V models to achieve state-of-the-art coherent storytelling"
X Link 2025-12-10T20:07Z 10.3K followers, 2684 engagements
"Sina Weibo AI Lab unveils VibeThinker-1.5B This 1.5B model challenges the scaling consensus achieving top-tier reasoning on math & coding. It surpasses 400x larger models on benchmarks like AIME25 trained at a fraction of the cost ($7800 vs $294K+)"
X Link 2025-11-12T04:12Z 10.3K followers, 2299 engagements
"OPPO AI Agent Team introduces O-Mem an Omni Memory System for LLM Agents It brings personalized long-horizon self-evolving capabilities to conversational AI by actively profiling users and supporting hierarchical retrieval for adaptive responses"
X Link 2025-11-30T12:10Z 10.3K followers, 14.4K engagements
"A new era of physics-aware video generation is here Introducing NewtonRewards a framework that enforces Newton's Laws in video diffusion models. It achieves physically plausible smooth and temporally coherent motions through verifiable rewards outperforming prior methods"
X Link 2025-12-03T00:20Z 10.3K followers, 11.7K engagements
"ByteDance's Nex-N1: Agentic Models Trained via a Unified Ecosystem Introducing Nex-N1 a groundbreaking platform for LLM agents It scales diverse complex interactive environments for effective policy learning outperforming SoTA models on SWE-bench & tau2"
X Link 2025-12-05T20:10Z 10.3K followers, 5859 engagements
"EditThinker learns to critique results refine instructions & repeat until satisfactory simulating human cognitive loops for better results. Paper:"
X Link 2025-12-08T04:16Z 10.3K followers, XXX engagements
"CAPO: New RL for stable generalized reasoning From Xiaomi & Tsinghua an adaptive curriculum uses positive-only advantage for stable foundations then introduces negative signals for discrimination. It boosts math & multimodal GUI reasoning. Plug-and-play with PPO GRPO RLOO & more"
X Link 2025-12-08T16:09Z 10.3K followers, 3623 engagements
"LongVT: Incentivizing "Thinking with Long Videos" This new framework for LMMs uses Multimodal Chain-of-Tool-Thought enabling global-to-local reasoning in long videos via native video cropping to tackle hallucinations and outperform baselines"
X Link 2025-12-02T04:15Z 10.3K followers, 1438 engagements
"ByteDance Seed introduces DAComp: A benchmark for data agents DAComp challenges AI agents across the full data intelligence lifecycle: from complex Data Engineering (multi-stage SQL pipelines) to open-ended Data Analysis. SOTA models struggle XX% success on DE tasks"
X Link 2025-12-05T16:10Z 10.3K followers, 1454 engagements
"Mistral just dropped Mistral Large X on Hugging Face Their new state-of-the-art multimodal Mixture-of-Experts model boasts 41B active params 675B total and a massive 256k context window for frontier performance"
X Link 2025-12-02T15:39Z 10.3K followers, 8141 engagements
"Kuaishou's Kling Team introduces MultiShotMaster This pioneering framework enables highly controllable multi-shot video generation. It tackles narrative complexity with flexible shot arrangement coherent transitions and custom control for subjects and backgrounds"
X Link 2025-12-03T12:11Z 10.3K followers, 1436 engagements
"Kuaishou Technology introduces ViDiC A new benchmark for MLLMs to caption fine-grained similarities & differences between video pairs. ViDiC-1K dataset: 1000 video pairs 4000+ checklist items across X categories (subject motion style & more)"
X Link 2025-12-04T08:09Z 10.3K followers, 2337 engagements
"Microsoft just launched VibeVoice Realtime on Hugging Face A lightweight streaming text-to-speech model generating initial audible speech in 300ms perfect for live data and LLM conversations"
X Link 2025-12-06T17:04Z 10.3K followers, 218K engagements
"Wikontic: Constructing Wikidata-aligned Ontology-Aware Knowledge Graphs This multi-stage pipeline leverages LLMs to build compact consistent knowledge graphs from raw text boosting reasoning without relying on auxiliary data"
X Link 2025-12-07T08:09Z 10.3K followers, 6128 engagements
"Top AI Papers of The Week (December 1-7): - - DeepSeek-V3.2 by @deepseek_ai: Pushing open LLM frontiers with incredible reasoning & agent performance - A Practical Guide to Code Intelligence covering agents and applications from companies like Microsoft ByteDance & Anthropic - LongVT: Incentivizing "Thinking with Long Videos" via Native Tool Calling for LMMs - Z-Image: An Efficient Image Generation Foundation Model that rivals commercial models - Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length - DAComp: Benchmarking Data Agents across the Full Data"
X Link 2025-12-07T14:20Z 10.3K followers, 1929 engagements
"ByteDance-Seed just released Adversarial Flow Models on Hugging Face These new generative models unify adversarial and flow models achieving a new SOTA FID of XXXX on ImageNet-256px with a single forward pass"
X Link 2025-12-08T12:46Z 10.3K followers, 12.1K engagements
"Alibaba's Tongyi Lab introduces Wan-Move for precise video motion control Generates high-fidelity 5-second 480p videos with control rivaling commercial tools by making original condition features motion-aware"
X Link 2025-12-10T04:15Z 10.3K followers, 8520 engagements
"Alibaba & Tsinghua just dropped Wan-Move A scalable framework bringing fine-grained point-level motion control to video generation. It produces 5-second 480p videos whose quality rivals commercial tools like Kling XXX Pro's Motion Brush"
X Link 2025-12-10T08:45Z 10.3K followers, 4622 engagements
"Visionary democratizes 3DGS research & deployment by offering a lightweight unified platform right in your browser It lowers barriers to reproduce compare & deploy cutting-edge 3D generative models. Paper: Project page:"
X Link 2025-12-10T12:10Z 10.3K followers, XXX engagements
"It achieves state-of-the-art information retention (86% on MINE-1) and creates KGs using 1000 output tokens. Get the full details on Hugging Face:"
X Link 2025-12-07T08:10Z 10.3K followers, XXX engagements
"ByteDance Seed unveils UniUGP for end-to-end autonomous driving. It's a unified framework that synergizes scene understanding future video generation and trajectory planning. Achieves SOTA in complex long-tail scenarios"
X Link 2025-12-11T04:17Z 10.3K followers, 1674 engagements