Research map / Reinforcement learning / Visual and representation RL
TRACED: Transition-aware Regret Approximation with Co-learnability for Environment Design
Geonwoo Cho, Jaegyun Im, Jihwan Lee, Hojun Yi, Sejin Kim, Sundong Kim
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 |
|---|---|---|---|
| add ATPL and Co-Learnability scoring to ACCEL (ablation) | ACCEL · BipedalWalker (6 held-out terrains, ablation) | mean return: 49.67 -> 89.95 | Improved |
| add Co-Learnability to TRACED without ATPL (ablation) | TRACED - CL · BipedalWalker (6 held-out terrains, ablation) | mean return: 77.93 -> 89.95 | Improved |
| add ATPL to TRACED without Co-Learnability (ablation) | TRACED - ATPL · BipedalWalker (6 held-out terrains, ablation) | mean return: 62.19 -> 89.95 | Improved |
| use alpha=1.0 instead of 100.0 (ablation) | TRACED · MiniGrid (12 held-out tasks, scaling ablation) | mean solved rate: 0.07 -> 0.7 | Improved |
| use alpha=1.0 instead of 0.0 (ablation) | TRACED · MiniGrid (12 held-out tasks, scaling ablation) | mean solved rate: 0.6 -> 0.7 | Improved |
| use beta=1.0 instead of 0.0 (ablation) | TRACED · MiniGrid (12 held-out tasks, scaling ablation) | mean solved rate: 0.5 -> 0.7 | Improved |
| use beta=1.0 instead of 100.0 (ablation) | TRACED · MiniGrid (12 held-out tasks, scaling ablation) | mean solved rate: 0.46 -> 0.7 | Improved |
| use TRACED instead of ACCEL at 20k updates | curriculum method · PerfectMazeLarge | solved rate (%): 20% -> 27% | Improved |
| use TRACED at 10k instead of ACCEL at 20k | curriculum method · PerfectMazeXL | solved rate (%): 12% -> 10% | No clear effect |
| use TRACED instead of ACCEL at 10k updates | curriculum method · MiniGrid | wall-clock training time (hours): 12.94 -> 13.78 | Hurt |
| use TRACED at 10k instead of ACCEL at 20k | curriculum method · MiniGrid | wall-clock training time (hours): 26.58 -> 13.78 | Improved |
| use TRACED at 10k instead of ACCEL at 20k | curriculum method · BipedalWalker | wall-clock training time (hours): 70.22 -> 35.64 | Improved |
Closest papers by meaning
- No Regrets: Investigating and Improving Regret Approximations for Curriculum Discovery (NeurIPS 2024 · similarity 0.93)
- Improving Regret Approximation for Unsupervised Dynamic Environment Generation (NeurIPS 2025 · similarity 0.92)
- Adversarial Environment Design via Regret-Guided Diffusion Models (NeurIPS 2024 · similarity 0.90)
- Marginal Benefit Driven RL Teacher for Unsupervised Environment Design (AAAI 2025 · similarity 0.90)
- Improving Environment Novelty Quantification for Effective Unsupervised Environment Design (NeurIPS 2024 · similarity 0.90)
- Refining Minimax Regret for Unsupervised Environment Design (ICML 2024 · 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.
