[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.]
@hu_yifei Yifei Hu
Yifei Hu posts on X about $75m, $108m, unlock, consistency the most. They currently have XXXXX followers and XX posts still getting attention that total XXX engagements in the last XX hours.
Engagements: XXX #
- X Week XXXXXX +168%
- X Month XXXXXX +32,306%
- X Months XXXXXXX -XXXX%
- X Year XXXXXXX -XX%
Mentions: X #
- X Week X +350%
- X Month X +350%
- X Months XX +10%
- X Year XX -XX%
Followers: XXXXX #
- X Week XXXXX +4.80%
- X Month XXXXX +6%
- X Months XXXXX +18%
- X Year XXXXX +41%
CreatorRank: XXXXXXXXX #
Social Influence #
Social category influence vc firms #153 finance XXXX%
Social topic influence $75m 4.35%, $108m 4.35%, unlock 4.35%, consistency 4.35%, ab XXXX%
Top accounts mentioned or mentioned by @reductoai @aditabrm @a16z @alibabaqwen @realmrfakename @jeffdean @swyx @venki @liu_eroteme @lukeigel @allenai @jeremyphoward @ekapil_kapil @m_att_dunn @ironredsandhive @mjamei @eugeneyan @mayasolos @rogerw0108 @ananthamoorthi_
Top Social Posts #
Top posts by engagements in the last XX hours
"@lukeigel @aditabrm @reductoai @a16z The second floor restroom will always welcome first floor folks"
X Link @hu_yifei 2025-10-14T16:37Z 3949 followers, XX engagements
"We spent a lot of our fundings on R&D. Recently our ML team figured out a reliable training approach allowing auto-regressive models to output 300+ bounding boxes in dense documents without shifting/skipping. Our grounding VLMs can detect table cells input fields checkboxes by combining their text understanding and grounding capabilities"
X Link @hu_yifei 2025-10-14T19:49Z 3952 followers, 69.9K engagements
"We have been testing LLM/VLM-as-judge for our new models. We found that certain models show higher position consistency than others. For example we ran the same test set (n=253) with Grok-4-fast-reasoning and GPT-5-thinking-high. For each model we ran the same prompt twice with different candidate order: - here is the question - here is the criteria - here is answer A/B - here is answer B/A - which one is better (CoT explanation + rating) In this case Grok-4-fast-reasoning gave very different eval scores when we re-ordered A and B. Disclaimer: grok-4-fast-reasoning is about 30x cheaper and 5x"
X Link @hu_yifei 2025-10-06T17:34Z 3945 followers, 1096 engagements
"@aditabrm @reductoai @a16z Let's settle for 64x B200s for now"
X Link @hu_yifei 2025-10-14T16:15Z 3949 followers, XXX engagements
"So if you were using Qwen3-8B for text only tasks you should probably switch to Qwen3-VL-8B for better text-only performance. In vLLM set --limit-mm-per-prompt to X. @Alibaba_Qwen any chance you guys release Qwen3-8B-2510 🙏"
X Link @hu_yifei 2025-10-15T18:33Z 3952 followers, 1047 engagements
"@a16z @reductoai Why so serious"
X Link @hu_yifei 2025-10-14T16:17Z 3949 followers, 2337 engagements
"@aditabrm @reductoai @a16z That's a lot of $$$ to burn on GPUs"
X Link @hu_yifei 2025-10-14T16:12Z 3951 followers, 2567 engagements