Research map / Reinforcement learning / Visual and representation RL
MTRL-CG: Multi-Task Reinforcement Learning Method with Spectral Clustering-Based Task Grouping
Wenjia Meng, Teng Zhang, Haoliang Sun, Yilong Yin
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 MTRL-CG task grouping | CARE · Meta-World MT10-Fixed | max success rate: 0.77 -> 0.97 | Improved |
| add MTRL-CG task grouping | MT-SAC · Meta-World MT10-Mixed | max success rate: 0.70 -> 0.87 | Improved |
| add MTRL-CG task grouping | CMTA · Meta-World MT10-Mixed | final success rate: 0.58 -> 0.53 | Hurt |
| add MTRL-CG task grouping | PaCo · Meta-World MT50-Fixed | final success rate: 0.52 -> 0.51 | No clear effect |
| add MTRL-CG task grouping | PaCo · Meta-World MT50-Mixed | final success rate: 0.62 -> 0.66 | Improved |
| add MTRL-CG task grouping | CMTA · Meta-World MT50-Fixed | final success rate: 0.52 -> 0.59 | Improved |
| use spectral affinity grouping instead of HOA grouping (ablation) | CARE · Meta-World MT10-Fixed | max success rate: 0.87 -> 0.97 | Improved |
| use spectral affinity grouping instead of random grouping (ablation) | CARE · Meta-World MT10-Mixed | final success rate: 0.47 -> 0.73 | Improved |
| use affinity-based grouping instead of no grouping (random grouping lowers it) (ablation) | CARE · Meta-World MT10-Mixed | final success rate: 0.69 -> 0.47 | Hurt |
| use 3 groups instead of 10 (ablation) | CARE+MTRL-CG · Meta-World MT10-Mixed | final success rate: 0.34 -> 0.73 | Improved |
| use 3 groups instead of 5 (ablation) | CARE+MTRL-CG · Meta-World MT10-Mixed | max success rate: 0.72 -> 0.87 | Improved |
| use 3 groups instead of a single shared policy (1 group) (ablation) | CARE+MTRL-CG · Meta-World MT10-Fixed | max success rate: 0.77 -> 0.97 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PaCo — PaCo: Parameter-Compositional Multi-task Reinforcement Learning (NeurIPS 2022) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Meta-Reinforcement Learning via Exploratory Task Clustering (AAAI 2024 · similarity 0.91)
- Multi-Task Reinforcement Learning with Mixture of Orthogonal Experts (ICLR 2024 · similarity 0.89)
- Mastering Massive Multi-Task Reinforcement Learning via Mixture-of-Expert Decision Transformer (ICML 2025 · similarity 0.88)
- Towards an Information Theoretic Framework of Context-Based Offline Meta-Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.87)
- QMP: Q-switch Mixture of Policies for Multi-Task Behavior Sharing (ICLR 2025 · 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.
