Research map / Reinforcement learning / Imitation learning
Few-Shot Bayesian Imitation Learning with Logical Program Policies
Tom Silver, Kelsey R. Allen, Alex K. Lew, Leslie Pack Kaelbling, Josh Tenenbaum
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 |
|---|---|---|---|
| Use the LPP structured policy class and Bayesian learning from one demonstration | BreakoutNoFrameskip-v4 · 10 held-out instances, 500 frames | reward: reward of 3 in all 10 cases | Improved |
| Use the LPP structured policy class and Bayesian learning from at most five demonstrations | Nim and Checkmate Tactic · held-out task instances | demonstrations needed for test performance 1.0: 20 -- 1,000x data efficiency gains | Improved |
| Search complete policies by vanilla program induction instead of learning Boolean combinations incrementally | VPI · all tasks other than Nim | test performance | Hurt |
| Replace Boolean learning with a neural network classifier over enumerated DSL features (ablation) | LPP · six strategy games | test performance | Hurt |
| Use an L1-regularized neural network classifier over enumerated DSL features (ablation) | LPP · six strategy games | test performance | Hurt |
| Replace the grammatical prior with a uniform prior (ablation) | LPP · six strategy games | test performance | Hurt |
| Replace the grammatical prior with a prior penalizing only the number of top-level programs (ablation) | LPP · six strategy games | test performance | Hurt |
| Assume no demonstration noise in the likelihood | LPP · Nim with noisy demonstrations | test performance | Hurt |
| Use too high a noise probability in the likelihood | LPP · Nim | test performance | Hurt |
| Use a very small but nonzero noise probability in the likelihood | LPP · Nim with noisy demonstrations | test performance | Improved |
Closest papers by meaning
- Uncertainty-Based Smooth Policy Regularisation for Reinforcement Learning with Few Demonstrations (NeurIPS 2025 · similarity 0.89)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · similarity 0.88)
- Causal Imitation for Markov Decision Processes: a Partial Identification Approach (NeurIPS 2024 · similarity 0.87)
- Agnostic Interactive Imitation Learning: New Theory and Practical Algorithms (ICML 2024 · similarity 0.87)
- Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration (ICLR 2025 · similarity 0.87)
- Prompted Policy Search: Reinforcement Learning through Linguistic and Numerical Reasoning in LLMs (NeurIPS 2025 · similarity 0.87)
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.
