1. X
  2. Xin Zhang | 张鑫
Log inSign up
Xin Zhang | 张鑫
210 posts
user avatar
Xin Zhang | 张鑫
@xinzhangai
NLP | LLM, PhD student. Living w/ Long-Covid.
Hong Kong
izhx.github.io
Joined August 2018
519
Following
262
Followers
RepliesRepliesMediaMedia

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • user avatar
    Xin Zhang | 张鑫
    @xinzhangai
    Jan 1
    Checkout Youtu-LLM ! 🚀🚀🚀 2B small SOTA model for agent tasks! 🤖 Re: Two labs affiliated to differenct BGs of Tencent
    user avatar
    Rosinality
    @rosinality
    Jan 1
    Report on building a small LLM, from pre-training to post-training, mostly focused on synthesizing agent trajectories. Maybe they will pursue agentic post-training further later. btw, what is the relationship between Tencent AI Lab and YouTu Lab?
    257
  • user avatar
    Xin Zhang | 张鑫
    @xinzhangai
    Nov 13, 2025
    🚀 ERNIE 5.0 is here! Native omini-modal, unified autoregressive architecture, ultra-sparse MoE. Now on web, app & API platform. As China’s AI pioneer, @Baidu_Inc once again delivers cutting-edge innovation! Looking forward to smaller-scale models (2B–8B) to build omni-embed!
    user avatar
    Baidu Inc.
    @Baidu_Inc
    Nov 13, 2025
    Here comes ERNIE 5.0 — our latest natively omni-modal foundational model. It excels in omni-modal understanding, creative writing, instruction following, and more. We will continue investing in and developing more cutting-edge models to push the boundaries of intelligence.
    208
  • user avatar
    Xin Zhang | 张鑫
    @xinzhangai
    Sep 10, 2025
    Impressive models! 🚀 Congrats!!! We now have new SOTA encoder backbone. Our mGTE-MLM-base outperforms XLM-R-base in 2024 summer (or jan by training), its a "semi-modern" encoder 😃. Glad to see it stands as one of only two baselines, alongside XLM-R.
    user avatar
    Orion Weller
    @orionweller
    Sep 9, 2025
    XLM-R has been SOTA for 6 years for multilingual encoders. That's an eternity in AI 🤯 Time for an upgrade. Introducing mmBERT: 2-4x faster than previous models ⚡ while even beating o3 and Gemini 2.5 Pro 🔥 + open models & training data - try it now! How did we do it? 🧵
    318
  • user avatar
    Xin Zhang | 张鑫
    @xinzhangai
    Sep 5, 2025
    New open-weight small embedding model surpass mGTE !! With Matryoshka Embedding 🪆
    user avatar
    Omar Sanseviero
    Google AI Studio
    @osanseviero
    Sep 4, 2025
    Introducing EmbeddingGemma🎉 🔥With only 308M params, this is the top open model under 500M 🌏Trained on 100+ languages 🪆Flexible embeddings (768 to 128 dims) with Matryoshka 🤗Works with your favorite open tools 🤏Runs with as little as 200MB developers.googleblog.com/en/introducing…
    280
  • user avatar
    Xin Zhang | 张鑫
    @xinzhangai
    Aug 18, 2025
    New 1.5B embedding and reranking models 🤩 !!! New choice between Qwen3-embedding-0.6B and 4B We release **Lychee-embed** and **Lychee-rerank**, based-on Qwen2.5-1.5B and our multi-stage training framework in COLM 2025 paper. #NLP #LLM #RAG #COLM2025
    18K