Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Pragmatically Learning from Pedagogical Demonstrations in Multi-Goal Environments
Hugo Caselles-Dupré, Olivier Sigaud, Mohamed CHETOUANI
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 both a pedagogical teacher and pragmatic learner instead of a naive teacher and literal learner | GANGSTR teacher-learner training · Fetch Block Stacking, 100 demos per goal | GIA x GRA: 0.15 -> 0.62 | Improved |
| Use both a pedagogical teacher and pragmatic learner instead of a naive teacher and literal learner | GANGSTR teacher-learner training · Fetch Block Stacking, 500 demos per goal | GIA x GRA: 0.62 -> 0.80 | Improved |
| Use both a pedagogical teacher and pragmatic learner instead of a naive teacher and literal learner | GANGSTR teacher-learner training · Fetch Block Stacking, 1000 demos per goal | GIA x GRA: 0.70 -> 0.82 | Improved |
| Train the teacher to infer its own goals and reward correctly inferred demonstrations | Naive teacher · Fetch Block Stacking teacher demonstrations | Own Goal Inference Accuracy: 83.4% ± 0.2 -> 95.6% ± 0.1 | Improved |
| Train the teacher to produce demonstrations it can disambiguate | Naive teacher · Fetch Block Stacking, 500 ambiguous situations | Ambiguity Score: 64% ± 2 -> 9% ± 1 | Improved |
| Train the learner to infer its own goals using BGI and reward correct inferences | Literal learner · Fetch Block Stacking | Own Goal Inference Accuracy: 12.4% relative difference | Improved |
| Replace BGI with a separately trained LSTM goal prediction network (ablation) | BGI goal inference · Fetch Block Stacking, pedagogical demos, 100 demos per goal | GIA: 69.2% -> 46.2% | Hurt |
| Replace BGI with a separately trained LSTM goal prediction network (ablation) | BGI goal inference · Fetch Block Stacking, pedagogical demos, 500 demos per goal | GIA: 82.2% -> 70.0% | Hurt |
| Replace BGI with a separately trained LSTM goal prediction network (ablation) | BGI goal inference · Fetch Block Stacking, naive demos, 100 demos per goal | GIA: 61.3% -> 38.4% | Hurt |
| Use demonstrations only and discard learner-collected experience (ablation) | Modified SQIL learner · Fetch Block Stacking, pedagogical demos, 100 demos per goal | GIA: 69.2±4.2 -> 5.3 ± 1.5 | Hurt |
| Add behavioral-cloning updates on demonstrations (ablation) | Modified SQIL learner · Fetch Block Stacking, pedagogical demos, 100 demos per goal | GIA: 69.2±4.2 -> 2.3 ± 1.8 | Hurt |
| Use original SQIL instead of modified SQIL (ablation) | Modified SQIL learner · Fetch Block Stacking, pedagogical demos, 100 demos per goal | GIA: 69.2±4.2 -> 59.1 ± 2.4 | Hurt |
Closest papers by meaning
- Showing versus doing: Teaching by demonstration (NeurIPS 2016 · similarity 0.88)
- Learning from Demonstrations via Capability-Aware Goal Sampling (NeurIPS 2025 · similarity 0.87)
- Reward Learning From Narrated Demonstrations (CVPR 2018 · similarity 0.87)
- Bayesian Inference of Temporal Task Specifications from Demonstrations (NeurIPS 2018 · similarity 0.87)
- Modeling Human Understanding of Complex Intentional Action with a Bayesian Nonparametric Subgoal Model (AAAI 2016 · similarity 0.86)
- GOALNET: Interleaving Neural Goal Predicate Inference with Classical Planning for Generalization in Robot Instruction Following (AAAI 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.
