Research map / Reinforcement learning / Visual and representation RL
TeachMyAgent: a Benchmark for Automatic Curriculum Learning in Deep RL
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 |
|---|---|---|---|
| use ALP-GMM or Covar-GMM without expert knowledge rather than relying on expert-informed curriculum methods | Stump Tracks challenge experiments · Stump Tracks | percentage of mastered test tasks | Mixed |
| provide an initial easy-task distribution to ADR | ADR · Stump Tracks challenges | percentage of mastered test tasks | Helped |
| provide an initial easy-task distribution to GoalGAN | GoalGAN · Stump Tracks challenges | percentage of mastered test tasks | Helped |
| use ADR on a rugged difficulty landscape | ADR · Stump Tracks, rugged difficulty | percentage of mastered test tasks | Hurt |
| use GoalGAN with a forgetting student | GoalGAN · Stump Tracks, forgetting student | percentage of mastered test tasks | Hurt |
| use SPDL when evaluation tasks are uniformly sampled over the task space | SPDL · Stump Tracks challenge experiments | percentage of mastered test tasks: 11/15 experiments | Hurt |
| use a Gaussian target distribution rather than a target matching the evaluation distribution | SPDL · Stump Tracks | percentage of mastered test tasks | Hurt |
| use ADR with a randomly selected initial task distribution in a rugged task space | ADR · Parkour | percentage of mastered test tasks | Hurt |
| use GoalGAN instead of expert-knowledge-free teachers | GoalGAN · Parkour | percentage of mastered test tasks | No clear effect |
| use SPDL or Setter-Solver | SPDL and Setter-Solver · Parkour | percentage of mastered test tasks | No clear effect |
| train with the climbing morphology rather than an easier-to-learn morphology | All teachers · Parkour, climbing morphology | percentage of mastered test tasks: at most 1% | Hurt |
Closest papers by meaning
- PORTAL: Automatic Curricula Generation for Multiagent Reinforcement Learning (AAAI 2024 · similarity 0.90)
- Unsupervised Training Sequence Design: Efficient and Generalizable Agent Training (AAAI 2024 · similarity 0.88)
- Online Reinforcement Learning-Based Pedagogical Planning for Narrative-Centered Learning Environments (AAAI 2024 · similarity 0.88)
- Curriculum Design for Trajectory-Constrained Agent: Compressing Chain-of-Thought Tokens in LLMs (NeurIPS 2025 · similarity 0.87)
- Learning Progress Driven Multi-Agent Curriculum (ICML 2025 · similarity 0.87)
- Diffusion-based Curriculum Reinforcement Learning (NeurIPS 2024 · 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.