<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;">
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
Dear all,</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<br>
</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
We are pleased to invite you to the next meeting of the VUML Seminar. Please see more details below the email:</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<br>
</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
Speaker: <b>Annie Wong</b>, PhD candidate at Leiden University</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
Title: State Design Matters: How Representations Shape Dynamic Reasoning in Large Language Models</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<br>
</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
The meeting will take place on <b>Monday 16 March 2026 at 10:00 in room NU-3A57</b>. We hope to see many of you there!</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<br>
</div>
<div style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
The organizers,</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<br>
</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
Zhao & Daniel</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<br>
</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
———————————————</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<span style="font-size: 12pt;"><b>Abstract</b>: </span>Large Language Models (LLMs) are increasingly used as decision-making agents in dynamic environments. However, a critical yet often overlooked factor in deploying such agents is how the state of the environment
is represented to the model. In this talk, based on our recent paper ‘State Design Matters: How Representations Shape Dynamic Reasoning in Large Language Models’, we investigate how state representation influences an LLM’s ability to reason and act sequentially.
Across multiple benchmarks, we systematically vary the representation of the agent’s state while keeping the underlying model fixed. We study three key design dimensions: granularity, comparing long-form context with concise trajectory summaries; structure,
contrasting natural language with symbolic or code-like encodings; and spatial grounding, evaluating text-only representations against representations augmented with images or textual maps.</div>
<div class="elementToProof" style="font-size: 12pt; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, Calibri, Helvetica, sans-serif;">
<div class="elementToProof" style="font-size: 12pt;"><b>Bio</b>: Annie Wong is a PhD candidate at Leiden University, jointly affiliated with the Natural Computing Group and the Reinforcement Learning Group. Her research investigates agents for sequential decision-making,
with a particular focus on reinforcement learning and evolutionary strategies. Alongside her academic work, she works as a Senior Data Scientist at Microsoft, where she applies AI and machine learning methods in industry.</div>
</div>
</body>
</html>