Research map / Large language models / Language, safety and interpretability
Building Social World Model with Large Language Models
Haofei Yu, Yining Zhao, Guanyu Lin, Jiaxuan You
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| scale world model backbone from 0.6B to 8B (ablation) | SWM · Kalshi (attr, posterior mode) | MASE: 0.884 -> 0.738 | Improved |
| scale world model backbone from 0.6B to 8B (ablation) | SWM · Kalshi (attr, posterior mode) | Corr: 0.290 -> 0.525 | Improved |
| scale world model backbone from 0.6B to 8B (ablation) | SWM · Polymarket (attr, posterior mode) | MASE: 0.977 -> 0.892 | Improved |
| lengthen look-back window from 1 to 16 (ablation) | SWM · Kalshi (attr, posterior mode) | MASE: 0.787 -> 0.738 | Improved |
| lengthen look-back window from 1 to 16 (ablation) | SWM · Kalshi (attr, posterior mode) | Corr: 0.430 -> 0.525 | Improved |
| scale prior attributor from 0.6B to 8B (ablation) | SWM prior attributor · Kalshi | median KL to oracle posterior: 2.10 -> 1.18 | Improved |
| scale prior attributor from 0.6B to 8B (ablation) | SWM prior attributor · Kalshi | Recall@3: 78% -> 88% | Improved |
| use sharper Qwen3.5-397B posterior instead of Qwen3-32B (ablation) | SWM training · Kalshi (attributed subset) | Corr: 0.357 -> 0.525 | Improved |
| use sharper Qwen3.5-397B posterior instead of Qwen3-32B (ablation) | SWM training · all-set coverage | attribution coverage: 47.8% -> 11.9% | Hurt |
| select event candidates with posterior instead of random (ablation) | SWM · Kalshi | MASE | Improved |
| replace prior attributor with oracle posterior at inference (ablation) | SWM · Kalshi | DA: 0.845 -> 0.894 | Improved |
| replace prior attributor with oracle posterior at inference (ablation) | SWM · Polymarket (attr) | DA: 0.685 -> 0.844 | Improved |
Closest papers by meaning
- MF-LLM: Simulating Population Decision Dynamics via a Mean-Field Large Language Model Framework (NeurIPS 2025 · similarity 0.87)
- Reaching a Consensus in Predictive Loops (NeurIPS 2026 · similarity 0.86)
- Position: LLM Social Simulations Are a Promising Research Method (ICML 2025 · similarity 0.86)
- Bias Amplification in Language Model Evolution: An Iterated Learning Perspective (NeurIPS 2024 · similarity 0.85)
- DEBATE: A Large-Scale Benchmark for Evaluating Opinion Dynamics in Role-Playing LLM Agents (NeurIPS 2026 · similarity 0.84)
- Social World Model-Augmented Mechanism Design Policy Learning (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
