←
Yali Du274 posts
Yali Du avatar

Yali Du

@yalidux

Researcher in reinforcement learning/multi-agent cooperation. @turinginst @KingsCollegeLon Ex at @AI_UCL @UCLCS

◷ Joined September 2016▧ 46 Photos and videos
Ming Jin avatar
⟳ Yali Du retweeted
Ming Jin@MingJin_AI·7h

Quick time update: Today's Agentic AI Frontier Seminar with Prof. Kaiqing Zhang starts at 12:30 PM ET / 9:30 AM PT (not 11 AM ET). apologies for the earlier typo, and see you on the same Zoom link shortly!

LLMs weren't trained for sequential decision-making—so when are LLM agents actually no-regret, and when do they fail? Excited to host Prof. Kaiqing Zhang (UMD) @KaiqingZhang at the Agentic AI Frontier Seminar! He’ll dive into decision-making agents through the lens of regret: • When off-the-shelf LLMs achieve vanishing regret (and where adversarial/variable losses break them) • RMFT: Regret-minimization fine-tuning without optimal action labels • RegPO: Step-level local regret credit assignment so good early actions aren't penalized for later mistakes—boosting web, household & robotics benchmarks 📅 Wed, Nov 18 | 11 AM ET / 8 AM PT 🔗 Zoom & info: agentic-ai-frontier-seminar.…

Post attachment
Yali Du avatar
Yali Du@yalidux·Nov 16

A wonderful trip to Hong Kong last October! 🇭🇰 Grateful to the Hong Kong Institute for Advanced Study for inviting me to the HKIAS International Conference on Multi-Agent Systems and AI at City University of Hong Kong. It was a rare chance to sit down with mathematicians, including Fields Medalist Pierre-Louis Lions, and discuss what AI means for mathematics at a moment when AI-generated proofs are making waves. I even got to bring home the handwritten notes from his talk! Along the way I visited colleagues at HKBU, PolyU, HKUST and CityUHK, with many stimulating conversations on multi-agent systems, LLM agents and AI for science, and the joy of catching up with old friends, Bo Han Gangyan Xu Sen Li Minming Li Chengqi Zhang. Hong Kong is a city full of energy and charm. Looking forward to coming back soon! 🌹 #MultiAgentSystems #AI #Mathematics #AIforMath #HongKong

Post attachmentPost attachmentPost attachmentPost attachment
Yali Du avatar
Yali Du@yalidux·Nov 12

5 days left to make a submission! Don't miss the opportunity to join the ELLIS UnConference 2026! @ELLISforEurope

Call for Papers: Multi-Agent AI for Scientific Discovery workshop at ELLIS UnConference 2026 We invite paper submissions to the workshop Multi-Agent AI for Scientific Discovery (yalidu.github.io/ellis-multi…), in Paris right before NeurIPS, December 8, 2026. Invited speakers include: Prof Samuel Kaski @samikaski (University of Manchester and Aalto University), Prof Jun Wang @seawan (University College London), Dr Zafeirios Fountas @zfountas (Huawei Technologies, London). Organised by Yali Du (King's College London) and Jun Wang (University College London). We welcome work on multi-agent AI systems for scientific discovery, including learning and coordination, shared knowledge, hypothesis generation and experimental design, human-AI collaboration, and evaluating when agent teams help. Applications across the sciences are welcome, such as life sciences and healthcare, chemistry and materials science, physics and astronomy, and earth and climate science. We invite double-blind non-archival submissions in either of the following categories: - Category A: original papers of up to 4 pages, using the workshop LaTeX template. Preliminary results and emerging ideas are welcome. - Category B: papers prepared for a top ML/AI venue (accepted, under review or previously submitted), in their original format and length. Selected submissions will be presented in the ELLIS UnConference poster session, and three selected contributions will be invited to give lightning talks. At least one author of each accepted paper must attend in person. Submission deadline: 18 November 2026, AoE (Anywhere on Earth). Author notification: 23 November 2026, AoE. Workshop date: 8 December 2026 Submission link: openreview.net/group?id=ELLI… Workshop website: yalidu.github.io/ellis-multi… For any questions, please reach out to: yali.du@kcl.ac.uk.

Yali Du avatar
Yali Du@yalidux·Nov 12

Had a wonderful day at the University of Liverpool @LivUni ! 🧪🤖 Thrilled to give a talk at the Department of Chemistry and to visit the Autonomous Chemistry Lab and Materials Innovation Factory with Prof Andy Cooper, Xenofon Evangelopoulos and Gabriella Pizzuto. It was great to see first-hand how robots and AI are reshaping the way chemistry is done, including the KUKA mobile robotic chemist featured in Nature. We had lots of fun discussions at the intersection of AI and chemistry, from multi-agent coordination in the lab to how AI agents can plan and drive scientific discovery. Many exciting directions to explore together! Huge thanks to Andy Cooper and Xenofon Evangelopoulos for the warm hospitality. 🙏 liverpool.ac.uk/materials-in… #AI #AutonomousLabs #MultiAgentSystems #AIforScience #Chemistry

Post attachmentPost attachmentPost attachmentPost attachment
Yali Du avatar
Yali Du@yalidux·Oct 28

Call for Papers: Multi-Agent AI for Scientific Discovery workshop at ELLIS UnConference 2026 We invite paper submissions to the workshop Multi-Agent AI for Scientific Discovery (yalidu.github.io/ellis-multi…), in Paris right before NeurIPS, December 8, 2026. Invited speakers include: Prof Samuel Kaski @samikaski (University of Manchester and Aalto University), Prof Jun Wang @seawan (University College London), Dr Zafeirios Fountas @zfountas (Huawei Technologies, London). Organised by Yali Du (King's College London) and Jun Wang (University College London). We welcome work on multi-agent AI systems for scientific discovery, including learning and coordination, shared knowledge, hypothesis generation and experimental design, human-AI collaboration, and evaluating when agent teams help. Applications across the sciences are welcome, such as life sciences and healthcare, chemistry and materials science, physics and astronomy, and earth and climate science. We invite double-blind non-archival submissions in either of the following categories: - Category A: original papers of up to 4 pages, using the workshop LaTeX template. Preliminary results and emerging ideas are welcome. - Category B: papers prepared for a top ML/AI venue (accepted, under review or previously submitted), in their original format and length. Selected submissions will be presented in the ELLIS UnConference poster session, and three selected contributions will be invited to give lightning talks. At least one author of each accepted paper must attend in person. Submission deadline: 18 November 2026, AoE (Anywhere on Earth). Author notification: 23 November 2026, AoE. Workshop date: 8 December 2026 Submission link: openreview.net/group?id=ELLI… Workshop website: yalidu.github.io/ellis-multi… For any questions, please reach out to: yali.du@kcl.ac.uk.

Yali Du avatar
Yali Du@yalidux·Sep 2

Proud moment as a supervisor! Today, Yudi Zhang defended her PhD at TU/e. I've been working with Yudi and Mykola Pechenizkiy @pechenizkiy for over four years. She is not only a talented student, but also a close friend of mine. Her PhD journey went from learning agents and interpretable decision making, to causal RL, and on to self-evolving decision-making systems. The thesis cover, front and back, is a lovely personal touch — designed by Yudi herself, weaving in her own interests and passions. Glad to see her thrive. She'll be joining TU Delft to work with Frans @faoliehoek. All the best for the next chapter! 愿你做个永远仰望星空的小女孩。 May you always be the girl who looks up at the stars.

Post attachmentPost attachmentPost attachmentPost attachment
Yali Du avatar
Yali Du@yalidux·Aug 31

AI Agent Frontier Seminar series is coming back at 2 Sep! Fantastic talk by @SharonYixuanLi at 11 AM ET! Register to get the link to attend! Organisers: @MingJin_AI @ShangdingG95714 @lifu_huang @ChenguangWang @yalidux

We are thrilled to kick off the Fall 2026 AI Agent Frontier Seminar this Wed, Sept 2 at 11 AM ET! 🚀 Prof. Sharon Li (UW-Madison, ICML 2026 Chair) @SharonYixuanLi will present: "Toward Reliable Deployment of LLM Agents: From Uncertainty Quantification to Progress Advantage." How can we tell if an LLM agent is quietly failing mid-episode during long-horizon tasks? She will explore agent Uncertainty Quantification (UQ) and how RL post-training yields a powerful implicit signal: the progress advantage. 🔗 Join us: virginiatech.zoom.us/j/87872… 📅 Full Schedule: agentic-ai-frontier-seminar.… Organizers: @ShangdingG95714 @yalidux @lifu_huang @ChenguangWang @MingJin_AI #LLMs #AIAgents #MachineLearning #TrustworthyAI #ReinforcementLearning

Yali Du avatar
Yali Du@yalidux·Aug 17

On Tue, 18 Aug, our paper “Causality-Inspired Spatial-Temporal Return Decomposition Approach for Multi-Agent Reinforcement Learning” will be presented at @RL_Conference 2026 in Montreal! Joint work with Yudi Zhang, Yali Du, @huang_biwei (Biwei Huang), @pechenizkiy (Mykola Pechenizkiy), and Meng Fang. 🎤 Talk 3 · 11:40 AM–12:30 PM 📌 Poster #58 · 3:00–6:00 PM

Post attachment
Yali Du avatar
Yali Du@yalidux·Jul 29

Congratulations to Dr Ziyan Wang @ZiyanWang98 , who has passed his PhD viva with no corrections! It has been a great pleasure to supervise Ziyan and see his work on safe and cooperative AI develop across ICML, NeurIPS, AAAI, AAMAS, TMLR, and COLM. Many thanks to his examiners, @casdewitt Christian Schroeder de Witt and Angelo Cangelosi, for their time and thoughtful feedback. He will soon join @MilindTambe_AI group at Harvard as a postdoc. A fantastic achievement and an exciting next step!

Post attachment
Yali Du avatar
Yali Du@yalidux·Jul 6

At ICML this Tuesday morning, three of our papers are postering in Hall A: 📍 #116 📍 #2004 📍 #2007 I’ll also be around — please stop by to chat with the presenting authors and discuss the work! 🌹🌹 @TristanTomilin Wei Liu @yulanhe

Post attachmentPost attachmentPost attachmentPost attachment
Yali Du avatar
Yali Du@yalidux·Jun 28

Before my summer travel, our group squeezed in a 20 km countryside hike 🌿🥾 By the end, everyone’s legs were sore and feet were complaining 😂. Time to work on the lab’s physical fitness as seriously as our research! Next stops: 📍 Jun 28 – Jul 1: Bay Area (Apple Park) 📍 Jul 2 – Jul 9: Seoul (KAIST, Seoul National University & ICML) 📍 Jul 9 – Jul 13: Beijing If you’re around, I’d be delighted to catch up and chat about human centric agentic systems, cooperative intelligence, multi agent safety or collaborations.

Post attachmentPost attachmentPost attachment
0:09
Yali Du avatar
Yali Du@yalidux·Jun 25

I was recently invited to @Google London office to attend the Google Education Leader Series. It was inspiring to see seven education-related demos from Googlers, as well as many creative applications of AI in education. These tools showed strong potential to support both teaching and studying. As students increasingly use AI as an interactive learning partner, I have been reflecting on an important question: what is the most irreplaceable role of an educator? Perhaps it is no longer only about transmitting knowledge. It is also about asking the right questions, helping students develop judgement, and supporting their long-term ability to learn independently. #AIforEducation @Google

Post attachmentPost attachmentPost attachment