Tuesday 6 October · COLM

Team: P1 human fidelity; P2 personas/populations; P3 interactions/emergence; P4 validation. Two sequential poster conversations per person where listed. All times local PDT. Linked titles open the source.

Time	Programme / count	Four-person assignment
08:45–09:00	Opening remarks	All four
09:00–10:00	Chelsea Finn — Lessons from Robots for Developing Reliable LLMs	Optional all: general reliability
10:00–11:00	Oral 1; CollabSkill 10:15, intent reasoning 10:30	P1/P4 supporting methods; P2/P3 reading
11:00–13:00	Poster session 1 140 total; 4 direct	P1: Mind the Sim2Real Gap in User Simulation for Agentic Tasks Imperial Ballroom #64; then The Hidden Puppet Master: Predicting Human Belief Change in Manipulative LLM Dialogues Imperial Ballroom #72 P2: InterviewSim: A Scalable Framework for Interview-Grounded Personality Simulation Imperial Ballroom #65 P3: Misalignment Contagion: Can a Misaligned Minority Shift Aligned Agents in Multi-Agent LLM Deliberation? Imperial Ballroom #69 P4: The Illusion of Stochasticity in LLMs Grand Ballroom #115
13:00–14:30	Lunch / notes / follow-up	All four; do not schedule another formal poster block
14:30–15:30	Jennifer Pan — Neutrality Without Neutral Models	All four: opinion/information systems
15:30–16:30	Oral 2 (supporting cognition/fairness)	P4 optional; no new direct simulation result assumed
16:30–18:30	Poster session 2 144 total; 7 direct	P1: Identity, Cooperation and Framing Effects within Groups of Real and Simulated Humans Imperial Ballroom #3 P2: Diversifying Multiple Generative Agents by Aligning with Human Populations Imperial Ballroom #47; then Escaping the Nash Trap: Structural Estimation and Alignment of Strategic Reasoning in Large Language Models Imperial Ballroom #46 P3: Simulating Organized Group Behavior: New Framework, Benchmark, and Analysis Imperial Ballroom #4; then Talk is Cheap, Communication is Hard: Dynamic Grounding Failures and Repair in Multi-Agent Negotiation Franciscan A #76 P4: Small Foundation Models of Human Cognition and Behaviour Imperial Ballroom #5; then Can AI Truly Represent Your Voice in Deliberations? A Comprehensive Study of Large-Scale Opinion Aggregation with LLMs Grand Ballroom #111

Scope: 856 unique main papers / six poster blocks; 29 confirmed directly relevant papers. All titles and 203 abstracts screened. Oral/poster repeats are counted once. This short guide is a selection, not the complete reading list.

Main schedule and board locations use saved official extracts; live freshness unconfirmed. Confirm onsite. Additional suggestions are optional, not extra commitments.

Wednesday 7 October · COLM

PDT · Four-person priority routes · Sequential stops within each poster block

Time	Programme / count	Four-person assignment
09:00–10:00	John Langford — Industrial Innovation and its implications	Optional all
10:00–11:00	Oral 3; Do Humans and LLMs Diverge in Belief Revision? Evidence from a Bayesian Analysis at 10:45	P1/P4: human-comparison method
11:00–13:00	Poster session 3 142 total; 4 direct	P1: Do Humans and LLMs Diverge in Belief Revision? Evidence from a Bayesian Analysis Grand Ballroom #144; then The Story Shapes the Agent: Narrative Priors in LLM Behavior Imperial Ballroom #51 P2: What Makes a Sale? Simulating End-to-End Seller-Buyer Retail Dynamics with LLM Agents Franciscan A #78 P3: Commitment To Cooperation With Self-Negotiated Contracts Franciscan C #96; then SOTOPIA-TOM: Evaluating Privacy and Information Management in Multi-Agent Interaction with Theory of Mind Franciscan C #91 P4: Compared to What? Baselines and Metrics for Counterfactual Prompting Imperial Ballroom #41; then CIDER: A Dataset of Contextual Disclosure Boundaries for Privacy Preference Alignment Franciscan B #90
13:00–14:30	Lunch / notes / follow-up	All four; do not schedule another formal poster block
14:30–15:30	Panel: James Evans, Jeremy Avigad, Rose Yu, Alison Gopnik	P1/P2 optional; topic not published
15:30–16:30	Oral 4 (no confirmed direct paper)	Optional / prepare afternoon questions
16:30–18:30	Poster session 4 143 total; 6 direct	P1: Multilingual Agent-Based World Modeling for Social Science Imperial Ballroom #6 P2: Agent Bazaar: Enabling Economic Alignment in Multi-Agent Marketplaces Imperial Ballroom #17 P3: Social World Models Imperial Ballroom #7 P4: UrbanLLMind: Scalable LLM-Powered Urban Mobility Simulation with Open-Weight Models Imperial Ballroom #16

Main schedule and board locations use saved official extracts; live freshness unconfirmed. Confirm onsite. Additional suggestions are optional, not extra commitments.

Thursday 8 October · COLM

PDT · Four-person priority routes · Sequential stops within each poster block

Time	Programme / count	Four-person assignment
09:00–10:00	Henry Farrell — Large Language Models are Culture Machines	All four: culture framing
10:00–11:00	Oral 5 (no confirmed direct simulation paper)	Optional; otherwise reading/follow-ups
11:00–13:00	Poster session 5 144 total; 3 direct	P1: Illusory Truth or Mere Exposure? Model-Dependent Repetition Effects in LLM-Based Social Media Simulations Grand Ballroom #115 P2: Attractor States Emerge in Multi-Turn LLM Conversations Imperial Ballroom #8 P3: SocialVeil: Probing Social Intelligence of Language Agents under Communication Barriers Grand Ballroom #140 P4: Illusory Truth or Mere Exposure? Model-Dependent Repetition Effects in LLM-Based Social Media Simulations Grand Ballroom #115
13:00–14:30	Lunch / notes / follow-up	All four; do not schedule another formal poster block
14:30–15:30	Jason Eisner — Is NLP System-Building a Machine Learning Problem?	Optional; relevance not inferred
15:30–16:30	Oral 6; What AI Benchmarks Actually Measure: Adapting Convergent and Discriminant Validity to Interrogate Fifty-Six AI Benchmarks at 16:00	P4: construct validity; poster begins at 16:40
16:40–18:30	Poster session 6 143 total; 5 direct	P1: High Volatility and Action Bias Distinguish LLMs from Humans in Group Coordination Imperial Ballroom #34; then RealUserSim: Bridging the Reality Gap in Agent Benchmarking via Grounded User Simulation Imperial Ballroom #40 P2: TIDES: A Longitudinal Bilingual Dataset for Modeling Multi-Party Social Dynamics Imperial Ballroom #30; then Beyond expert users: agents should help users construct preferences, not just elicit them Franciscan C #99 P3: MoltNet: Understanding Social Behavior of AI Agents in the Agent-Native MoltBook Imperial Ballroom #31; then INSIDE the Student’s Mind: Jointly Modeling Latent Reasoning and Action in LLM Student Simulators Grand Ballroom #103 P4: What AI Benchmarks Actually Measure: Adapting Convergent and Discriminant Validity to Interrogate Fifty-Six AI Benchmarks Grand Ballroom G-6-117; then In-Situ Behavioral Evaluation for LLM Fairness, Not Standardized-Test Scores Imperial Ballroom #39

Main schedule and board locations use saved official extracts; live freshness unconfirmed. Confirm onsite. Additional suggestions are optional, not extra commitments.

Friday 9 October · parallel workshops

PDT · P1/P2: Social Sim · P3/P4: Agent Behavior. Stay in assigned workshop; no transfer required.

Time band	Social Sim, P1 and P2	Agent Behavior, P3 and P4
08:30 to 09:00	No scheduled session.	08:30 to 09:00. Opening remarks.
09:00 to 10:00	09:00 to 09:10. Opening remarks. 09:10 to 09:40. Noah Goodman, Person Simulation 09:40 to 10:15. Spotlights.	09:00 to 09:30. Natasha Jaques, AI Safety is a Multi-Agent Problem 09:30 to 10:00. Sophia Kazinnik, AI Agents as Economic Actors
10:00 to 11:30	Spotlights finish at 10:15. 10:15 to 10:45. Zhongyu Wei, Personalized Foundation Model for Social Simulation 10:45 to 11:00. Coffee. 11:00 to 11:30. Slava Jankin, Can Machines Negotiate a Government? LLM Agents, Coalition Formation, and the Validity Problem in Computational Social Science	10:00 to 11:30. Poster Session I P3: action-grounded evaluation and social dynamics. P4: asset markets and psychometrics. Named routes below.
11:30 to 12:30	11:30 to 12:30. Poster session P1/P2: two primary conversations each.	11:30 to 13:00. Lunch.
12:30 to 13:30	12:30 to 13:30. Lunch.	Lunch ends at 13:00. 13:00 to 14:00. Spotlight talks

Time band	Social Sim, P1 and P2	Agent Behavior, P3 and P4
13:30 to 14:00	13:30 to 14:00. Marwa Abdulhai	Spotlights continue until 14:00.
14:00 to 14:30	Logan Cross, Concordia: Generative Agent-Based Modelling for Social Science	Diyi Yang, Delegation, Reliance, and Oversight in Human-Agent Collaboration
14:30 to 15:30	14:30 to 15:30. Breakout session	14:30 to 15:00. Break. 15:00 to 15:30. Kawin Ethayarajh, The World Worlds
15:30 to 16:00	15:30 to 15:45. Coffee. 15:45 to 17:00. Panel discussion	15:30 to 16:00. Colin Camerer, Modelling Humans: In Silico & In Mice
16:00 to 17:10	Panel continues until 17:00. 17:00 to 17:10. Closing remarks.	16:00 to 17:30. Poster Session II P3: consensus and belief dynamics. P4: empirical validation. Named routes below.
17:10 to 18:00	Social Sim has finished.	Posters finish at 17:30. 17:30 to 18:00. Closing remarks.

18 workshops investigated: six complete rosters, five partial, seven unavailable. These two are priority choices, not exhaustive coverage. Talk titles/topics marked unpublished remain unknown. Poster/spotlight repeats are not additional papers.

Friday · priority paper conversations

Social Sim: P1/P2, 11:30–12:30. Agent Behavior: P3/P4, I 10:00–11:30; II 16:00–17:30. Two stops per person in each assigned block.

Owner	Linked paper / relevance	Why attend
P1	BehaviorBench: Benchmarking Foundation Models for Behavioral Science Tasks direct confirmed	Behavioral-science benchmarks and distributional evaluation
P1	A Model of Multi-turn Human Persuadability Using Probabilistic Belief Tracing direct confirmed	Human multi-turn belief trajectories and Bayesian latent beliefs
P2	Distinguishing Governance Effects from Prompting Artifacts in LLM Pricing Simulations direct candidate	Economic governance versus prompting confound
P2	Everyone Conforms, No One Believes: Pluralistic Ignorance in LLM Agent Populations direct candidate	Pluralistic ignorance and collective conformity
P3_I	Evaluating Generative Agents with Actions Grounded in Socially Distributed Task Environments using Incognita supporting adjacent	action-grounded social-agent evaluation
P3_I	What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates supporting adjacent	emergent social structure/objectives
P4_I	Persona Design and Bubble Formation in LLM-Populated Experimental Asset Markets direct candidate	persona sensitivity and emergent market behavior
P4_I	Rethinking Psychometric Evaluation of LLMs: When and Why Self-Reports Predict Behavior supporting adjacent	Big Five versus behavior-specific TPB; context/persona confounds across 11 models
P3_II	Inherited or Formed? The Provenance of Consensus in LLM Agent Societies direct candidate	emergent versus inherited consensus validity
P3_II	Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation direct candidate	direct inspectable opinion-dynamics tool
P4_II	Augmented Hypothesis Testing with Persona-Based LLM Simulations direct confirmed	statistically valid use of fallible persona predictions; four real datasets
P4_II	Can AI Agents Simulate A/B Test Outcomes? A Validation Framework for Agentic Experimentation direct confirmed	simulated RCT error decomposition and calibration against 67 historical tests

ABSTRACT-confirmed direct work is distinguished from title candidates and adjacent work. Social Sim individual poster attendance is not separately mapped. Ask every author: what human data grounds this model, which held-out test validates it, and does it beat a simpler baseline?