Research map / Reinforcement learning / Visual and representation RL
CLUTR: Curriculum Learning via Unsupervised Task Representation Learning
Abdus Salam Azad, Izzeddin Gur, Jasper Emhoff, Nathaniel Alexis, Aleksandra Faust, Pieter Abbeel, Ion Stoica
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 CLUTR with flexible regret instead of PAIRED with flexible regret | PAIRED on CarRacing · F1 benchmark | return: 82% | Improved |
| Use CLUTR with standard regret instead of PAIRED with standard regret | PAIRED on CarRacing · F1 benchmark | return: 10.6X | Improved |
| Use CLUTR with flexible regret instead of PAIRED with flexible regret | PAIRED on MiniGrid · 16 unseen navigation tasks | solve rate: 35% | Improved |
| Use CLUTR with standard regret instead of PAIRED with standard regret | PAIRED on MiniGrid · 16 unseen navigation tasks | solve rate: 45% | Improved |
| Keep the pretrained task decoder fixed during curriculum learning (ablation) | CLUTR on CarRacing · CarRacing | return: 29% performance drop | Improved |
| Keep the pretrained task decoder fixed during curriculum learning (ablation) | CLUTR with flexible regret · CarRacing | return: 10% | Improved |
| Sort task sequences when training the VAE (ablation) | CLUTR on CarRacing · CarRacing | return: 31% performance drop | Improved |
| Use flexible regret instead of standard regret | PAIRED on CarRacing · F1 benchmark | return: 257 ± 16 vs 26 ± 19 | Improved |
| Use flexible regret instead of standard regret | CLUTR on CarRacing · F1 benchmark | return: 468 ± 21 vs 276 ± 24 | Improved |
Closest papers by meaning
- Beyond Fixed Tasks: Unsupervised Environment Design for Task-Level Pairs (AAAI 2026 · similarity 0.90)
- Curriculum Reinforcement Learning using Optimal Transport via Gradual Domain Adaptation (NeurIPS 2022 · similarity 0.89)
- Unsupervised Training Sequence Design: Efficient and Generalizable Agent Training (AAAI 2024 · similarity 0.88)
- No Regrets: Investigating and Improving Regret Approximations for Curriculum Discovery (NeurIPS 2024 · similarity 0.88)
- Diffusion-based Curriculum Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Improving Environment Novelty Quantification for Effective Unsupervised Environment Design (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.
