Skip to content

Commit 3054dbf

Browse files
RSTZZZRSTZZZ
authored
Move talk details for May 5, 2026
Move the entry for the talk on 'Consistently Simulating Human Personas with Multi-Turn Reinforcement Learning', including presenter details and abstract.
1 parent 416ab8b commit 3054dbf

1 file changed

Lines changed: 31 additions & 31 deletions

File tree

‎stamina/index.html‎

Lines changed: 31 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -190,37 +190,7 @@ <h4>[DATE Y/M/D]</h4>
190190
--><!-- END TALK TEMPLATE -->
191191

192192
<br>
193-
194-
<h4>2026/05/05</h4>
195-
<li>
196-
<b><a href="[PAPER LINK]">Consistently Simulating Human Personas with Multi-Turn Reinforcement Learning</a></b>
197-
<br>
198-
Presenter: <u><a href="https://abdulhaim.github.io/" target="_blank" rel="noopener noreferrer">Marwa Abdulhai</a></u>, UC Berkeley AI Research (BAIR) Lab
199-
<a class="btn btn-info btn-xs" data-toggle="collapse" href="#20260505-bio" role="button" aria-expanded="false">
200-
Speaker Bio
201-
</a>
202-
<div class="collapse" id="20260505-bio">
203-
<div class="card card-body">
204-
Marwa Abdulhai is a PhD candidate at UC Berkeley advised by Sergey Levine. Her research focuses on enabling AI agents to better understand people and their interactions to build both safe and more AI capable systems. This includes improving the performance of existing large language models (LLMs) for multi-turn dialogue interactions, understanding how to protect against deception in AI systems, and exploring how AI can serve as a useful tool for social science research. Her research has been supported by the Quad Fellowship, AI Policy Hub, Open AI Research, and Cooperative AI PhD Fellowship.
205-
</div>
206-
</div>
207-
<br>
208-
<!-- <a href="[RECORDING LINK - ADD AFTER TALK]"><img src="https://img.shields.io/badge/Youtube-Recording-orange"></a> -->
209-
<!-- <a href="[PAPER LINK]"><img src="https://img.shields.io/badge/Paper-link-important"></a> -->
210-
<!-- <a href="[GITHUB_LINK]"><img src="https://img.shields.io/badge/Github-link-lightgrey"></a> -->
211-
<!-- <a href="[SLIDES_LINK]"><img src="https://img.shields.io/badge/Talk-Slides-blue"></a> -->
212-
<a class="btn btn-primary btn-xs" data-toggle="collapse" href="#20260505-abstract" role="button" aria-expanded="false">
213-
Abstract
214-
</a>
215-
<div class="collapse" id="20260505-abstract">
216-
<div class="card card-body">
217-
Large Language Models (LLMs) are increasingly used to simulate human users in interactive settings such as therapy, education, and social role-play. While these simulations enable scalable training and evaluation of AI agents, off-the-shelf LLMs often drift from their assigned personas, contradict earlier statements, or abandon role-appropriate behavior. We introduce a unified framework for evaluating and improving consistency in LLM-generated dialogue with multi-turn RL, reducing inconsistency by over 55%, resulting in more coherent and trustworthy simulated users.
218-
</div>
219-
</div>
220-
</li>
221-
222-
<br/>
223-
193+
224194
<h4>2026/05/19</h4>
225195
<li>
226196
<b><a href="[PAPER LINK]">The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models</a></b>
@@ -273,6 +243,36 @@ <h3 style="text-align:center">Past Talks (<a href="https://www.youtube.com/@Comp
273243

274244
<h4 style="text-align:center; margin-top:30px;">Spring 2026</h4>
275245

246+
<h4>2026/05/05</h4>
247+
<li>
248+
<b><a href="[PAPER LINK]">Consistently Simulating Human Personas with Multi-Turn Reinforcement Learning</a></b>
249+
<br>
250+
Presenter: <u><a href="https://abdulhaim.github.io/" target="_blank" rel="noopener noreferrer">Marwa Abdulhai</a></u>, UC Berkeley AI Research (BAIR) Lab
251+
<a class="btn btn-info btn-xs" data-toggle="collapse" href="#20260505-bio" role="button" aria-expanded="false">
252+
Speaker Bio
253+
</a>
254+
<div class="collapse" id="20260505-bio">
255+
<div class="card card-body">
256+
Marwa Abdulhai is a PhD candidate at UC Berkeley advised by Sergey Levine. Her research focuses on enabling AI agents to better understand people and their interactions to build both safe and more AI capable systems. This includes improving the performance of existing large language models (LLMs) for multi-turn dialogue interactions, understanding how to protect against deception in AI systems, and exploring how AI can serve as a useful tool for social science research. Her research has been supported by the Quad Fellowship, AI Policy Hub, Open AI Research, and Cooperative AI PhD Fellowship.
257+
</div>
258+
</div>
259+
<br>
260+
<a href="https://youtu.be/4sA8Xe6mCZQ"><img src="https://img.shields.io/badge/Youtube-Recording-orange"></a>
261+
<!-- <a href="[PAPER LINK]"><img src="https://img.shields.io/badge/Paper-link-important"></a> -->
262+
<!-- <a href="[GITHUB_LINK]"><img src="https://img.shields.io/badge/Github-link-lightgrey"></a> -->
263+
<!-- <a href="[SLIDES_LINK]"><img src="https://img.shields.io/badge/Talk-Slides-blue"></a> -->
264+
<a class="btn btn-primary btn-xs" data-toggle="collapse" href="#20260505-abstract" role="button" aria-expanded="false">
265+
Abstract
266+
</a>
267+
<div class="collapse" id="20260505-abstract">
268+
<div class="card card-body">
269+
Large Language Models (LLMs) are increasingly used to simulate human users in interactive settings such as therapy, education, and social role-play. While these simulations enable scalable training and evaluation of AI agents, off-the-shelf LLMs often drift from their assigned personas, contradict earlier statements, or abandon role-appropriate behavior. We introduce a unified framework for evaluating and improving consistency in LLM-generated dialogue with multi-turn RL, reducing inconsistency by over 55%, resulting in more coherent and trustworthy simulated users.
270+
</div>
271+
</div>
272+
</li>
273+
274+
<br/>
275+
276276
<h4>2026/04/28</h4>
277277
<li>
278278
<b><a href="[PAPER LINK]">From Social Networks to Sensemaking Networks</a></b>

0 commit comments

Comments
 (0)