Xiaoyin Chen

PhD student @Mila_Quebec, advised by Yoshua Bengio.

[Alexandre @alexpiche_]

In-flight weight updates have gone from a “weird trick” to a must to train LLMs with RL in the last few weeks. If you want to understand the on-policy and throughput benefits here’s the CoLM talk @DBahdanau and I gave: https://t.co/p3KMZLFg4l

[Sanxing Chen @sanxing_chen]

Most RL for LLMs today is single-step optimization on a given state (e.g., an instruction), which is essentially a bandit setup. But to learn a meta-policy that can solve various bandit problems via in-context trial and error, you need true multi-turn RL over a long horizon. So, can RL & SFT teach LLMs a meta-bandit policy to explore in-context? 🤔

The regret-based benchmarks screamed YES! But … real story is more complex. We discovered a surprising phenomenon “When Greedy Wins.”
(1/5) 🧵

[Jiarui Lu @jiarlu]

ICML'25 | Aligning Protein Conformation Ensemble Generation with Physical Feedback

We bake physical model's energy into protein structure prediction models (AF3) to make its ensemble generation powerful!😎 (1/5)
🔗arXiv: https://t.co/ZMxYm6iKBT
🐙Github: https://t.co/fZbbVoyqdL

[NewInML @ NeurIPS 2026 @NewInML]

New to ML research? Never published at ICML? Don't miss this!

Check out the New in ML workshop at ICML 2025 — no rejections, detailed feedback, awards, and ICML tickets for selected authors.

Deadline: June 10 (AoE)
Submit: https://t.co/xNiccKTelq
Info: https://t.co/1dBY6bnGji

[Hongyi James Cai @Hongyicai2002]

🚀Excited to share our new paper: How Much Backtracking is Enough?

🤔 How many backtracks should your LLM learn to reason better?
Turns out: the harder the task, the more backtracking you need!

[Minsu Kim @minsuuukim]

Bringing hierarchical structure for search-based inference in LLM.

Sharing our recent work "Search-Based Correction of Reasoning Chains for LMs". Work done with Jean-Pierre Falet, Oliver Richard, Xiaoyin Chen , @JainMoksh , @SungjinAhn_ , @sungsoo_ahn_, @Yoshua_Bengio

https://t.co/sUHt4PePUh

[Jiarui Lu @jiarlu]

ICLR'25 | Structure Language Models for Protein Conformation Generation
Keywords: language modeling | discrete diffusion | PLM fine-tuning
🔗arXiv: https://t.co/gBczsWAhYq
🐱🐙Github: https://t.co/Xym04P3cIt
📍Fri | 25 Apr. 10am | Hall 3 + Hall 2B
Happy to chat if you're in 🇸🇬!!

[Tianyu Zhang @tianyu_zh]

🚀A game-changer for open-access multimodal AI! BigDocs is paving the way for transparent, accountable, and innovative document reasoning and code generation. Check it out! 💡👏