Research map / Reinforcement learning / Imitation learning
Guiding Policies with Language via Meta-Learning
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 |
|---|---|---|---|
| add iterative language corrections to instruction following | GPL · Multi-room object manipulation test | completion rate: 0.066 -> 0.82 | Helped |
| add iterative language corrections to instruction following | GPL · Object relocation test | completion rate: 0.65 -> 0.90 | Helped |
| use interactive corrections instead of the initial instruction alone | GPL · Multi-room object manipulation test | completion rate: 0.075 -> 0.82 | Helped |
| use interactive corrections instead of the initial instruction alone | GPL · Object relocation test | completion rate: 0.64 -> 0.90 | Helped |
| add the initial instruction (ablation) | GPL · Multi-room object manipulation test, 5 corrections | completion rate: 0.79 -> 0.82 | Helped |
| add the previous trajectory to correction context (ablation) | GPL · Multi-room object manipulation test, 5 corrections | completion rate: 0.77 -> 0.82 | Helped |
| retain all previous corrections instead of only the immediate correction (ablation) | GPL · Multi-room object manipulation test, 5 corrections | completion rate: 0.63 -> 0.82 | Helped |
| provide more than the 5 corrections used during training | GPL · Multi-room object manipulation test | completion rate: 0.82 -> 0.86 | Helped |
| provide more than the 5 corrections used during training | GPL · Object relocation test | completion rate: 0.90 -> 0.95 | Helped |
| use directional rather than fully specified corrections | GPL · Multi-room object manipulation test, 5 corrections | completion rate: 0.56 vs 0.606 | Hurt |
| use binary yes/no corrections | GPL · Multi-room object manipulation test, 5 corrections | completion rate: 0.09 vs 0.606 | Hurt |
| generalize to objects held out during training | GPL · Multi-room object manipulation test, unseen objects | completion rate: 0.82 -> 0.75 | Mixed |
Closest papers by meaning
- Text-Aware Diffusion for Policy Learning (NeurIPS 2024 · similarity 0.89)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 2025 · similarity 0.89)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- Learning with Language-Guided State Abstractions (ICLR 2024 · similarity 0.88)
- RLZero: Direct Policy Inference from Language Without In-Domain Supervision (NeurIPS 2025 · similarity 0.88)
- Learning Agentic Policy from Action Guidance (NeurIPS 2026 · similarity 0.88)
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.