Research map / Reinforcement learning / Visual and representation RL
Multi-task Batch Reinforcement Learning with Metric Learning
Jiachen Li, Quan Vuong, Shuang Liu, Minghua Liu, Kamil Ciosek, Henrik Christensen, Hao Su
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 |
|---|---|---|---|
| Initialize SAC from the MBML multi-task policy instead of random initialization | SAC · HumanoidDir-M unseen tasks | sample efficiency: 85K vs 350K; 76% improvement | Improved |
| Initialize SAC from the MBML multi-task policy instead of random initialization | SAC · five MuJoCo task distributions | convergence speed: up to 80% improvement | Improved |
| Use transition relabelling and the proposed triplet loss rather than neither component (ablation) | MBML · unseen task distributions | average episode returns | Improved |
| Remove transition relabelling while retaining a triplet loss with mined context-set examples (ablation) | MBML · unseen task distributions | average episode returns | Hurt |
| Remove the triplet loss while retaining relabelled transitions (ablation) | MBML · unseen task distributions | average episode returns | Hurt |
| Remove both transition relabelling and the triplet loss (ablation) | MBML · WalkerParam | average episode returns | No clear effect |
| Use a learned reward function for relabelling instead of the ground-truth reward function (ablation) | MBML · UmazeGoal-M | average episode returns | No clear effect |
| Use a reward prediction ensemble for relabelling (ablation) | MBML initialization for SAC · unseen task distributions | convergence performance | Improved |
Closest papers by meaning
- Task-Aware Virtual Training: Enhancing Generalization in Meta-Reinforcement Learning for Out-of-Distribution Tasks (ICML 2025 · similarity 0.87)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.87)
- Provable Benefits of Multi-task RL under Non-Markovian Decision Making Processes (ICLR 2024 · similarity 0.87)
- Hard Tasks First: Multi-Task Reinforcement Learning Through Task Scheduling (ICML 2024 · similarity 0.86)
- Decomposed Prompt Decision Transformer for Efficient Unseen Task Generalization (NeurIPS 2024 · similarity 0.86)
- DaCe-DT: Data-Centric Offline Multi-Task Reinforcement Learning via Adaptive Prompts and Trajectory Correction for Heterogeneous Tasks (NeurIPS 2026 · similarity 0.86)
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.
