Research map / Reinforcement learning / Imitation learning
Showing versus doing: Teaching by demonstration
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| teaching by showing instead of doing (goal-based grid) | human demonstrators · Experiment 1 goal-based grid | disambiguating (showing) trajectories per goal trial: 1.82 vs 1.05 | Helped |
| teaching by showing instead of doing (feature-based reward) | human demonstrators · Experiment 2 colored-tile grid | entropy of colored-tile visits: 0.50 vs 0.39 | Helped |
| teaching by showing instead of doing (feature-based reward) | human demonstrators · Experiment 2 colored-tile grid | proportion of colored-tile visits after first colored visit: 0.87 vs 0.82 | Helped |
| teaching by showing instead of doing (feature-based reward) | human demonstrators · Experiment 2 colored-tile grid | state revisits (looping): 1.38 vs 0.10 | Helped |
| teaching by showing instead of doing (feature-based reward) | human demonstrators · Experiment 2 colored-tile grid | total rewards won | No clear effect |
| training MLIRL on showing trajectories instead of doing trajectories | MLIRL with human and model trajectories · Experiment 2 colored-tile grid | learned reward weight distance from true reward | Helped |
Closest papers by meaning
- Making AI Policies Transparent to Humans through Demonstrations (AAAI 2024 · similarity 0.87)
- Inverse Reinforcement Learning by Estimating Expertise of Demonstrators (AAAI 2025 · similarity 0.86)
- Learning Utilities from Demonstrations in Markov Decision Processes (ICML 2025 · similarity 0.85)
- Imitation Learning from Observation with Automatic Discount Scheduling (ICLR 2024 · similarity 0.85)
- Learning from Demonstrations via Capability-Aware Goal Sampling (NeurIPS 2025 · similarity 0.85)
- Rethinking Inverse Reinforcement Learning: from Data Alignment to Task Alignment (NeurIPS 2024 · similarity 0.85)
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.