Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Online Bayesian Goal Inference for Boundedly Rational Planning Agents
Tan Zhi-Xuan, Jordyn Mann, Tom Silver, Josh Tenenbaum, Vikash K. Mansinghka
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 |
|---|---|---|---|
| replace unbiased BIRL with SIPS | BIRL (unbiased) · Doors, Keys & Gems | Top-1 accuracy at Q2: 0.33 -> 0.74 | Improved |
| replace unbiased BIRL with SIPS | BIRL (unbiased) · Intrusion Detection | Top-1 accuracy at Q2: 0.05 -> 0.87 | Improved |
| replace unbiased BIRL with SIPS | BIRL (unbiased) · Block Words | average cost per timestep (s): 63.6 -> 4.15 | Improved |
| replace unbiased BIRL with SIPS | BIRL (unbiased) · Taxi | Top-1 accuracy at Q3: 0.92 -> 0.67 | Mixed |
| replace oracle BIRL with SIPS | BIRL (oracle) · Intrusion Detection | Top-1 accuracy at Q2: 1.00 -> 0.87 | Hurt |
| replace BIRL (oracle) with SIPS | BIRL (oracle) · Human goal inference trajectories | correlation with human inferences: 0.51 -> 0.89 | Improved |
| increase mismatch: lower continuation probability q (0.95 -> 0.8) in data (ablation) | SIPS · Doors, Keys, Gems | Q3 top-1 accuracy: 0.73 -> 0.53 | Hurt |
| use goal-count heuristic data instead of h_add (ablation) | SIPS · Block Words | Q3 top-1 accuracy: 0.87 -> 0.43 | Hurt |
| data from more optimal maze-distance heuristic instead of Manhattan (ablation) | SIPS · Doors, Keys, Gems | Q3 top-1 accuracy: 0.73 -> 0.90 | Improved |
| data from lower search noise gamma (0.1 -> 0.02) (ablation) | SIPS · Doors, Keys, Gems | Q3 top-1 accuracy: 0.73 -> 0.77 | Improved |
| data from human subjects instead of model-generated (ablation) | SIPS · Doors, Keys, Gems | Q3 top-1 accuracy: 0.73 -> 0.79 | No clear effect |
Closest papers by meaning
- Reflect-then-Plan: Offline Model-Based Planning through a Doubly Bayesian Lens (ICML 2025 · similarity 0.87)
- Modeling Boundedly Rational Agents with Latent Inference Budgets (ICLR 2024 · similarity 0.87)
- A Bayesian Approach to Online Planning (ICML 2024 · similarity 0.86)
- Pre-Trained Multi-Goal Transformers with Prompt Optimization for Efficient Online Adaptation (NeurIPS 2024 · similarity 0.86)
- Bayesian Exploration Networks (ICML 2024 · similarity 0.86)
- PcLast: Discovering Plannable Continuous Latent States (ICML 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.
