RT @omarsar0: Deep Research Agents with Test-Time Diffusion. Google keeps pushing on diffusion. This time, they apply diffusion to deep re….

#AI #LLM #DeepResearch #TestTimeScaling #Diffusion.

Authors: Rujun Han, Yanfei C., Zoey CuiZhu, Lesly Miculicich, Guan Sun, Yuanjun Bi, Weiming Wen, Hui Wan, Chunfeng Wen, Solène Maître, George Lee, Vishy Tirumalashetty, Yuan (Emily) Xue, Zizhao Zhang, Salem Haykal, Burak Gokturk, Tomas Pfister, and Chen-Yu Lee.

This work marks a significant step toward creating more effective AI research companions capable of tackling complex questions across diverse industry domains, including finance, biomedical, recreation, and technology.

Our method outperforms existing deep research agents across a wide array of benchmarks that require intensive search and multi-hop reasoning for long-form report generation.

much like a diffusion model sharpens an image. This draft-centric approach, combined with a self-evolutionary algorithm, makes the report writing process more timely and coherent, significantly reducing information loss during iterative searches.

Instead of a linear writing, our method re-frames research report generation as a "𝙙𝙞𝙛𝙛𝙪𝙨𝙞𝙤𝙣 𝙥𝙧𝙤𝙘𝙚𝙨𝙨." It begins with a "𝘯𝘰𝘪𝘴𝘺" initial draft, which is then iteratively refined through a "𝘥𝘦𝘯𝘰𝘪𝘴𝘪𝘯𝘨" process using externally retrieved information—

Thrilled to introduce "𝗗𝗲𝗲𝗽 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿 𝘄𝗶𝘁𝗵 𝗧𝗲𝘀𝘁-𝗧𝗶𝗺𝗲 𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻," a new deep research agent designed to mimic the iterative nature of human research, complete with cycles of planning, drafting, and revision. 🚀🚀.

RT @aistats_conf: The #AISTATS 2025 Test of Time Award goes to . 🥁 . Chen-Yu Lee, Saining Xie, Patrick Gallagher, Zhengyou Zhang, Zhuow….

Thank you to the AISTATS committee, my fantastic co-authors @sainingxie, Patrick Gallagher, Zhengyou Zhang, and my PhD advisor Zhuowen Tu for this incredible journey!. Paper: Award slides and recording (will be available later):

Feeling immensely proud that our paper "Deeply-Supervised Nets" has received a Test of Time Award at AISTATS 2025! It's incredibly rewarding to see the continued relevance and impact of this work, originally published back in 2015.

RT @shangbinfeng: 📌 Grab the models, define your goal, and let Heterogeneous Swarms find multi-LLM systems for you!. 📄 Paper: https://t.co/….

RT @cyjustinchen: Happy to share that RevThink has been accepted to #NAACL2025 main conference! 🎉We also release the code and data 👇🧵. RevT….

RT @cyjustinchen: 🚨 Reverse Thinking Makes LLMs Stronger Reasoners. We can often reason from a problem to a solution and also in reverse to….

RT @WendaXu2: Searching for the ultimate LLM Knowledge distillation! Want one that excels in both task-specific and task-agnostic settings.….

RT @shangbinfeng: 👀 How to find a better adapted model?.✨ Let the models find it for you!. 👉🏻 Introducing Model Swarms, multiple LLM expert….

RT @GoogleAI: Speculative RAG splits Retrieval Augmented Generation tasks into two separate steps of drafting followed by verification to p….

RT @zlwang_cs: Interested in tabular understanding with LLMs? Check out our Chain-of-Table at #ICLR2024 ! The poster will be at Hall B #73….

RT @cydhsieh: Why LLMs lost in the middle❓.💡LLMs exhibit U-shape positional attention bias that dominates their generation behavior (often….

Big kudos to the fantastic team for making this work possible! @IHung_Hsu, @ZifengWang315, Long T. Le, Lesly Miculicich, @VioletNPeng, @chl260, @tomaspfister 6/n