Research map / Large language models / AI and society
Modeling Others' Minds as Code
Kunal Jha, Aydan Yuenan Huang, Eric Ye, Natasha Jaques, Max Kleiman-Weiner
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 |
|---|---|---|---|
| model agent as synthesized Python program with Bayesian reweighting instead of naive LLM prompting | NLLM · Construction scripted, multi-step | accuracy: 0.310 -> 0.479 | Improved |
| use ROTE instead of NLLM on human gameplay | NLLM · Construction human, multi-step | accuracy: 0.151 -> 0.296 | Improved |
| use ROTE instead of NLLM on Partnr | NLLM · Partnr single-step | accuracy: 0.113 -> 0.537 | Improved |
| add two-stage observation parsing (ablation) | ROTE · Partnr single-step | accuracy: 0.537 -> 0.160 | Hurt |
| add two-stage observation parsing (ablation) | ROTE · Construction scripted and human | accuracy: 0.479 -> 0.522 scripted | No clear effect |
| enforce strict FSM structure (severe) instead of light (ablation) | ROTE · Partnr single-step | accuracy: 0.537 -> 0.440 | Hurt |
| use moderate structure instead of light (ablation) | ROTE · Construction human | accuracy: 0.296 -> 0.310 | Improved |
| use SMC with rejuvenation instead of importance sampling (ablation) | ROTE · small hypothesis budgets | early accuracy | Mixed |
| vary top-k pruning (1, 10, 30) (ablation) | ROTE · Construction | accuracy | No clear effect |
Closest papers by meaning
- OdysSim: Building Foundation Models for Human Behavior Simulation (NeurIPS 2026 · similarity 0.88)
- CogBench: a large language model walks into a psychology lab (ICML 2024 · similarity 0.87)
- AutoToM: Scaling Model-based Mental Inference via Automated Agent Modeling (NeurIPS 2025 · similarity 0.87)
- Towards Real-world Human Behavior Simulation: Benchmarking Large Language Models on Long-horizon, Cross-scenario, Heterogeneous Behavior Traces (NeurIPS 2026 · similarity 0.86)
- Towards Reliable Code-as-Policies: A Neuro-Symbolic Framework for Embodied Task Planning (NeurIPS 2025 · similarity 0.86)
- Beyond Single Stationary Policies: Meta-Task Players as Naturally Superior Collaborators (NeurIPS 2024 · similarity 0.86)
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.
