Updates

✨Very much looking forward to attending one of my favorite conferences—the Midwest Speech and Language Days, generously hosted this year by NLP@UIUC. As ACL and other large conferences continue to grow, I find myself gravitating toward NLP events where it’s easier to truly

🌴🌅Two of our Agent papers are accepted to ACL 2026 (@aclmeeting)!

1️⃣ TT-SI: Self-Improving LLM Agents with Test-Time Training: https://t.co/xPuspINtjo
2️⃣ Current Agent Fails to Leverage World Model as Tool for Foresight: https://t.co/Q17GIt8LEi

Both of our papers are part of

Very excited to virtually attend #EACL2026!
I'll be giving an oral presentation on our work:

Beyond Sample-Level Feedback: Using Reference-Level Feedback to Guide Data Synthesis
🔗 https://t.co/ShgY1PBraX
🗓️ March 25 | 17:30–19:00 UTC

I'm also serving as a session chair for

🎉 Excited to share that our survey "Must Read: A Comprehensive Survey of Computational Persuasion" has been accepted to ACM Computing Surveys!

📄 https://t.co/s7FMs68B1G

Interdisciplinary research has driven major scientific breakthroughs. We show that LLMs can identify interdisciplinary insights that present fresh perspectives and improve creative research ideation. https://t.co/sMffOaCygl

Check out Idea-Catalyst!
📄Paper: https://t.co/G9w2eMM13A
💻Code: https://t.co/CQ6g4yhNxv
🤗Data: https://t.co/xyoNYhPeh6
🌐 Blogpost: https://t.co/5KiqQQj4K7

Thanks to my amazing collaborators @ShuhaibMehri @dilekhakkanitur @convai_uiuc @dmguiuc !

researchers from U of Illinois created:

> Idea-Catalyst

unlike an ai scientist that runs experiments and tries to reach a scientific conclusion

Idea-Catalyst analyzes cross-disciplinary ideas and tries to help scientists discover novel angles for their research

code below 👇

This is a great Cross-RL (or Dual-RL) study where the generator and solver reinforce each other for zero-shot agentic LLM building. https://t.co/3fZcAFA5Tc

Huge thanks to incredible team effort @qiancheng1231, @jonashuebotter, @hengjinlp, @dilekhakkanitur, @tur_gokhan!

Useful Links:

📄 Paper: https://t.co/AN5ozPHzzB
🌐 Project Page: https://t.co/yaqtMUW0eK
💻 Code: https://t.co/gJDOgAn89E
🤗 Models: https://t.co/G3VzguX6JL
📈

Can LLMs self-evolve into general-purpose tool-calling agents without any external data?

Yes. Introducing "Tool-R0: Self-Evolving LLM Agents for Tool-Learning from Zero Data"! ♾️

Initialized from the same LLM, Tool-R0 co-evolves a Generator and Solver through self-play RL;

Working on speech and language processing in Midwest? Submit your abstract and join us at UIUC mid April! https://t.co/AcGYzvvMaK

🇧🇷🎉 Accepted to #ICLR_2026 — REIN: Conversational Error Recovery with Reasoning Inception

* Paper Link: https://t.co/NOTi8sjeIe

Excited to give a talk @GoogleResearch tomorrow on user-centered conversational agents!👩‍💻🔄🤖

I'll cover two directions:
(1) LLMs that reliably simulate users during multi-turn conversations
(2) Agents that continuously learn user preferences & improve collaboration across