NeurIPS 2025 Workshop on Multi-Turn Interactions in Large Language Models (original) (raw)

About The Workshop

The field of AI is entering a new era of interaction, profoundly shaped by the capabilities of Large Language Models (LLMs). While multi-turn interaction has been a long-standing pursuit in AI—from dialogue systems to multi-agent coordination—the advent of LLMs has radically transformed this landscape. These models now engage in complex, long-horizon interactions, process diverse data, and make crucial decisions in dynamic, human-centric scenarios.

This leap forward, however, brings forth critical new research questions and challenges that demand immediate attention:

The Workshop on Multi-Turn Interactions in LLMs is designed to be the central forum for addressing these pivotal questions. We invite researchers to contribute to defining the next generation of interactive AI, tackling these core challenges, and charting the course for future advancements in AI reasoning and planning. This workshop will concentrate on key areas where the extended use of LLMs presents both new challenges and opportunities, serving as a platform to discuss and refine methods for future improvements and evaluation for practical LLM use cases.

(1) Multi-Turn Settings and Tasks:

Exploring diverse multi-turn interaction paradigms including human-AI, AI-AI, and AI-environment interactions. We welcome research on new multi-turn tasks, position papers on emerging interaction paradigms, and studies on complex scenarios like web agents, tool usage, simulations, and collaborative multi-agent systems. This includes GUI agents, conversational AI, interactive planning, and other long-horizon interactive tasks.

(2) Multi-Turn Frameworks and Algorithms:

Novel methods and frameworks for multi-turn interactions, including various reinforcement learning approaches (PPO, GRPO, etc.), agent architectures, training pipelines, and algorithmic innovations. We seek research on addressing sparse rewards, effective credit assignment, improving training stability, rollout efficiency, and developing new RL methods specifically designed for long-horizon interactive settings with LLMs.

(3) Multi-Turn Evaluation:

Long-horizon evaluation methods that assess consistency, stability, strategic ability, and performance degradation over extended interactions. This includes measuring and predicting performance on complex multi-turn tasks, identifying accumulating errors or unexpected behaviors, and creating comprehensive test environments. We encourage work building upon existing benchmarks like GAIA, TravelPlanner, τ-Bench, and ColBench, as well as developing new evaluation paradigms.

(4) Multi-Turn Challenges:

Addressing critical challenges in extended interactions, with a focus on maintaining alignment and safety over long-term interactions. This includes ensuring LLMs remain aligned with human values, maintaining consistent model personas, accurately tracking and adapting to users' changing goals, ensuring personalization does not compromise safety or fairness, and preventing advanced jailbreaking or hidden goal changes. We welcome research on coherence, personalization, trust, and other emerging challenges in multi-turn settings.

The Workshop on Multi-Turn Interactions in LLMs @ NeurIPS 2025 invites submissions on the development of novel architectures, algorithms, theoretical analyses, empirical studies, and applications in multi-turn interactions with LLMs. Submissions must present original, unpublished research.

Key Dates

Deadlines are strict and will not be extended under any circumstances. All deadlines follow the Anywhere on Earth (AoE) timezone.

Submission Site

Scope

We welcome contributions across a broad spectrum of topics related to our themes. Accepted papers will be presented as posters, with a subset selected for oral presentations. The workshop will take place in person at NeurIPS 2025, with virtual participation options to be confirmed.

Submission Guidelines

Formatting Requirements

Submissions must be in English and follow the NeurIPS 2025 Workshop LaTeX Template.

Papers must be submitted as a single PDF file:

Submissions exceeding the page limit will be desk rejected.

Anonymity

The workshop follows a double-blind review process. Submissions must be anonymized by removing author names, affiliations, and acknowledgments. Prior work should be cited in the third person. Identifying information, including in supplementary materials, must be omitted.

Dual Submission and Non-Archival Policy

NeurIPS and ICLR submissions are welcome to submit to our workshop. You can submit your work if it's currently under review at other venues. However, if your paper got accepted by other venues (e.g., NeurIPS) after submission, you have to withdraw from our workshop. Papers that are accepted after the submission deadline will not need to withdraw from our workshop, since the workshop is not archival. (That was a mistake previously, sorry!)

Transparency

By submitting to the workshop, authors agree that for all accepted papers, the original submission, reviews, and meta-reviews will be made publicly available on OpenReview.

Contact

Email at multiturn-interactions-organizers@googlegroups.com

Yu Su
Yu Su

Ohio State University

Yu Su
Yu Su

Ohio State University

Time (PDT) Session Speaker Talk Title
08:20 – 08:30 Opening Remarks Organizers
08:30 – 09:00 Invited Talk 1 Natasha Jaques (UW & Google DeepMind) Multi-turn Reinforcement Learning for LLMs
09:00 – 09:30 Oral Presentation + Lightning Talk 1 2 Oral + 5 Spotlight Talks
09:30 – 10:00 Invited Talk 2 Diyi Yang (Stanford) Human-AI Collaboration Across Turns
10:00 – 10:30 Invited Talk 3 Dawn Song (UC Berkeley) AI Agent Safety in the Wild
10:30 – 11:00 Invited Talk 4 Peter Henderson (Princeton) Building Better Rules and Finding Shallow Alignment in AI Agents
11:00 – 12:00 Poster Session 1
12:00 – 13:00 Lunch Break
13:00 – 13:30 Invited Talk 5 Jason Weston (Meta FAIR) Longer-Horizon Task Training
13:30 – 14:00 Invited Talk 6 Will Brown (Prime Intellect) Verifiers and Environment Scaling for RL
14:00 – 14:30 Oral Presentation + Lightning Talk 2 2 Oral + 5 Spotlight Talks
14:30 – 15:00 Invited Talk 7 Yu Su (Ohio State University) Computer Use: Modern Moravec’s Paradox
15:00 – 15:45 Panel Discussion Jason Weston, Natasha Jaques, Will Brown, Eric Wang, Yu Su (Coordinator: Weiyan Shi)
15:45 – 16:45 Poster Session 2
16:45 – 17:00 Paper Award & Closing Remarks Organizers

Oral Presentations

Spotlight Presentations

Click to expand spotlight papers (9 papers)

Poster Presentations

Click to expand poster papers (110 papers)

The Multi-Turn Interactions in LLMs (MTI-LLM) Workshop at NeurIPS 2025 is pleased to announce limited financial support for student authors of accepted workshop papers.

Thanks to our generous sponsors — Meta AI and Orby AI (now Uniphore) — we are offering registration reimbursement grants to ensure broader student participation in the NeurIPS community.

💡 Eligibility & Coverage

💬 Reimbursement Policy

Reimbursements will be processed after NeurIPS 2025 upon submission of a valid payment receipt for the student registration.

This ensures funds are distributed fairly to participants who attend and present at the workshop.

🗓️ Key Dates

🎯 Priority Considerations

Preference will be given to:

Registration Grant Recipients

Bo Liu
Bo Liu

National University of Singapore / Meta

Hanxu Hu
Hanxu Hu

University of Zurich / Microsoft Research Asia

Prime Intellect

Prime Intellect