Research map / Reinforcement learning / Imitation learning
Information Directed Reward Learning for Reinforcement Learning
David Lindner, Matteo Turchetta, Sebastian Tschiatschek, Kamil Ciosek, Andreas Krause
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 |
|---|---|---|---|
| update candidate policies 20 times instead of uniform sampling | IDRL · Swimmer-Corridor | regret after 20 queries: 11.8 -> 2.4 | Improved |
| replace IGR with IDRL (20 updates) | IDRL · Ant-Corridor | regret after 20 queries: 17.2 -> 2.2 | Improved |
| replace EIR with IDRL (20 updates) | IDRL · Ant-Corridor | regret after 20 queries: 17.5 -> 2.2 | Improved |
| reduce candidate policy updates from 20 to 4 (ablation) | IDRL · Swimmer-Corridor | regret after 20 queries: 2.4 -> 2.8 | Hurt |
| reduce candidate policy updates from 20 to 1 (ablation) | IDRL · Ant-Corridor | regret after 20 queries: 2.2 -> 12 | Hurt |
| reduce candidate policy updates from 20 to 1 (ablation) | IDRL · Swimmer-Corridor | runtime (hours): 40 -> 20 | Improved |
| reduce candidate policy updates from 20 to 1 (ablation) | IDRL · Ant-Corridor | runtime (hours): 40 -> 10 | Improved |
| single-update IDRL versus baselines (ablation) | IDRL (1 update) · Swimmer-Corridor | regret after 20 queries: 11.8 -> 7.6 | Improved |
| select queries by IDRL instead of pure information gain (IGR) | IDRL · Driver | regret | Improved |
| select queries by IDRL instead of EPD and IGR | IDRL · Gridworlds | regret | Improved |
| IDRL instead of baselines | IDRL · Swimmer-Corridor | cosine similarity of learned reward | No clear effect |
| IDRL instead of uniform/IGR | IDRL · MuJoCo Deep RL, 1400 comparisons | normalized score | Improved |
Closest papers by meaning
- Reinforcement Learning with Lookahead Information (NeurIPS 2024 · similarity 0.89)
- Efficient Preference-Based Reinforcement Learning: Randomized Exploration meets Experimental Design (NeurIPS 2025 · similarity 0.88)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.88)
- Exploration-Driven Policy Optimization in RLHF: Theoretical Insights on Efficient Data Utilization (ICML 2024 · similarity 0.88)
- An Information Theoretic Approach to Interaction-Grounded Learning (ICML 2024 · similarity 0.87)
- The Distributional Reward Critic Framework for Reinforcement Learning Under Perturbed Rewards (AAAI 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.
