✨Introducing ECHO, the newest in-the-wild image generation benchmark! You’ve seen new image models and new use cases discussed on social media, but old benchmarks don’t test them! We distilled this qualitative discussion into a structured benchmark. 🔗 echo-bench.github.io

🚀 End the GPU Cost Crisis Today!!! Headache with LLMs lock a whole GPU but leave capacity idle? Frustrated by your cluster's low utilization? We launch kvcached, the first library for elastic GPU sharing across LLMs. 🔗 github.com/ovg-project/kv… 🧵👇 Why it matters:

Humans handle dynamic situations easily, what about models? Turns out, they break in three distinct ways: ⛔ Force Stop → Reasoning leakage (won’t stop) ⚡️ Speedup → Panic (rushed answers) ❓ Info Updates → Self-doubt (reject updates) 👉Check out dynamic-lm.github.io

Had some really interesting discoveries recently: If a model performs extremely stable on one benchmark. Let's say a model is always getting 62% on SWEBench no matter what prompts or scaffold you used. It DOES NOT mean that the model is robust. It actually means that the model

The Sky’s Fun Committee, representing the ppl of sky, just dropped the new lab theme: ⚫️💖 Black Pink x Halloween 🎃🦇 We have: - Gru & the minions - kpop ??? 🫰😉

📢 ViDoRe V3, our new multimodal retrieval benchmark for enterprise use cases, is finally here! It focuses on real-world applied RAG scenarios using high-quality human-verified data. huggingface.co/blog/QuentinJG… 🧵(1/N)

Semantic search improves our agent's accuracy across all frontier models, especially in large codebases where grep alone falls short. Learn more about our results and how we trained an embedding model for retrieving code.

Tired: Grepping — O(n) Wired: Semantic search — O(log n) Index once. Query forever. Long turbopuffer 🐡

Turn your laptop into a powerful RAG system! LEANN can index and search through millions of documents while using 97% less storage than traditional solutions without accuracy loss. LEANN achieves this through graph-based selective recomputation with high-degree preserving

Turn your laptop into a powerful RAG system! LEANN can index and search through millions of documents while using 97% less storage than traditional solutions without accuracy loss. 100% Open Source

Thanks for sharing Sumanth , we will keep shipping new features! Let’s go LEANN, we just crossed 3000 GitHub stars last month.

The best open-source deep-research model I’ve seen so far. Tons of eye-opening insights on RLVR, real web-search grounding, and how LLMs actually use data. If you care about retrieval, reasoning, or agentic systems, this is a must-read.

Congrats to the team — incredible work! 👏 Interestingly, we built LEANN MCP in July (the first open-source semantic-search MCP for Claude Code), and we also saw: 🚀 50% token savings ⚡ 30% lower latency Repo👇 github.com/yichuan-w/LEAN…

Same situation. We released and kept RL going, it’s still giving gains😛

Alright so since state-of-the-art semantic search MCP is trending again, please let me introduce you to MCPyLate again 😇 x.com/antoine_chaffi… It probably is less powerful than Mixedbread’s search and have higher storage cost than LEANN, but it let’s you use any ColBERT

Yeah, super exciting to see Claude boosted with MS MARCO via latent-space search. Now it’s our turn — time to bring multi-vector retrieval to code 🚀 LEANN is already the first Claude-powered Code MCP, and CC+MultiVector+LEANN MCP (or even FastPlaid MCP Antoine Chaffin👀) is on

Hi NeurIPS Conference, I broke three bones but I still want to go to neurips. But seems like the conference is not very accessible🥲

grep is multimodal now. performs better then apple photos and gives you and your agent perfect search. just run npm install -g @mixedbread/mgrep

1/n 🚀 Introducing SkyRL-Agent, a framework for efficient RL agent training. ⚡ 1.55× faster async rollout dispatch 🛠 Lightweight tool + task integration 🔄 Backend-agnostic (SkyRL-train / VeRL / Tinker) 🏆 Used to train SA-SWE-32B, improving Qwen3-32B from 24.4% → 39.4%