🔥 Excited to share our latest work: WebWatcher 🕵️‍♂️ An open-source multimodal agent that achieves new SOTA on multiple challenging vision-language (VL) deep research benchmarks — outperforming GPT-4o & Gemini! Paper: arxiv.org/abs/2508.05748 Code: github.com/Alibaba-NLP/We…

Huggingface daily papers #1 Paper of the day! WebWatcher: Open-source multimodal agent that crushes VL research benchmarks, beating GPT-4o & Gemini! Introducing BrowseComp-VL: First open-source VL benchmark for deep web research! > Dynamic tool calls (search/browse/OCR/code)

Thrilled to open-source WebWatcher: our vision-language deep research agent from @Alibaba_NLP! Available in 7B & 32B parameter scales for the community. Achieving SOTA on the toughest VQA benchmarks: • HLE-VL: 13.6% (vs GPT-4o's 9.8%) • BrowseComp-VL: 27.0% (2x GPT-4o!) •

🎉 New Preprint Alert! 🎉 How can we improve Video-LLM's video understanding inspired by human's task-aware information filtering and cognitive load purification? 🤔 --"Temporal Visual Screening for Video-LLMs"

I was lucky to work in both China and the US LLM labs, and I've been thinking this for a while. The current values of pretraining are indeed different: US labs be like: - lots of GPUs and much larger flops run - Treating stabilities more seriously, and could not tolerate spikes

Accepted as NeurIPS2025 Spotlight! Existing large multimodal models (LMMs) have very poor visual understanding and reasoning over part-level attributes and affordances (only 5.9% gIoU). We developed novel part-centric LMMs to address these challenges arxiv.org/pdf/2505.20759

when you give up on this nebulous idea and illusion of prestige, you will finally find peace and freedom. submit to TMLR and JMLR.

🏗️ ARE: scaling up agent environments and evaluations In the LLM+RL era, evals and envs are the bottleneck Happy to release Gaia2, an extensible benchmark for agents aiming to reduce the sim2real gap + ARE, the platform in which Gaia2 is built Enjoy evaluating your agents! 👇

🚀 Introducing UserRL: a new framework to train agents that truly assist users through proactive interaction, not just chase static benchmarking scores. 📄 Paper: arxiv.org/pdf/2509.19736 💻 Code: github.com/SalesforceAIRe…

The decision notification letters have been sent! 🎉 We sincerely thank all authors and reviewers for their valuable contributions to this workshop. Kudos to our organizing committee, advisors, and support team for their incredible efforts: Guohao Li 🐫 May Fung (hiring postdocs) Qingyun Wang

See many of you (and your brilliant work in scaling agent environments) in San Diego! 🤩🎉 #NeurIPS2025

We are excited to announce 4 outstanding papers 🏆🏆🏆🏆 --> 🧵

We are very excited to welcome Dr. Ruhi Sarikaya from Alexa AI, Amazon to give a talk “Path to Artificial General Intelligence: Past, Present, and Future” at our UIUC CS Colloquium this Wed, 3:30 PM CT, in SC2405 and via Zoom: calendars.illinois.edu/detail/2654?ev…

We hear increasing discussion about aligning LLM with “diverse human values.” But what’s the actual price of pluralism? 🧮 In our #NeurIPS2025 paper (with Shawn Im), we move this debate from the philosophical to the measurable — presenting the first theoretical scaling law

Great keynote by the wonderful Heng Ji! Now I feel like I need to move into NLP+Science. I'm sure that we will see much more work in this direction thanks to her powerful advocating and her impactful work. EMNLP 2025 #EMNLP2025

#EMNLP Keynote by Heng Ji: No more Processing. Time to Discover! AI for Science is just so exciting! Let us make LLMs discover like true scientists: Observe → Think → Propose and Verify (A pity to miss the talk. Photo from May Fung (hiring postdocs) EMNLP 2025 )

So it seems Anthropic just rediscovered CodeAct 😄 arxiv.org/abs/2402.01030

Happy to be with many of my academic grandchildren from HKUST, and my PhD student Cheng Qian Cheng Qian @ EMNLP2025 . So proud of Prof. May Fung May Fung (hiring postdocs) and the awesome RenAI lab she has built: renai-lab.github.io

Diffusion language models are making a splash (again)! To learn more about this fascinating topic, check out ⏩ my video tutorial (and references within): youtu.be/8BTOoc0yDVA ⏩discrete diffusion reading group: Discrete Diffusion Reading Group

🎉 Congratulations to all #EMNLP2025 award winners 🎉 Starting with the ✨Best Paper award ✨: "Infini-gram mini: Exact n-gram Search at the Internet Scale with FM-Index" by Hao Xu, Jiacheng Liu, Yejin Choi, Noah A. Smith, and Hannaneh Hajishirzi aclanthology.org/2025.emnlp-mai… 1/n