Research map / Reinforcement learning / Visual and representation RL
Knowledge Transfer in Multi-Task Deep Reinforcement Learning for Continuous Control
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 |
|---|---|---|---|
| KTM-DRL vs ideal (K single-task TD3 agents) | KTM-DRL · Benchmark A | max average reward: 8743 -> 10348 on HCSmallTorso | Helped |
| KTM-DRL vs TD3-MT baseline | multi-task TD3 · Benchmark B | max average reward: 324 -> 5836 on Ant | Helped |
| KTM-DRL vs gradient surgery | multi-task agent · Benchmark B | max average reward: 2310 -> 10921 on HalfCheetah | Helped |
| add offline knowledge transfer stage (ablation) | KTM-DRL · Benchmark B | performance ratio to ideal: 0.8977 -> 1.0437 on Hopper | Helped |
| add online learning stage (ablation) | KTM-DRL · Benchmark B | performance ratio to ideal: 0.9035 -> 0.9745 on Ant | Helped |
| add teacher guidance term during online learning (ablation) | KTM-DRL · Benchmark B | performance ratio to ideal: 0.9872 -> 1.0112 on Walker2d | Helped |
| add critic's own gradient on top of teacher-only guidance online (ablation) | KTM-DRL · Benchmark B | performance ratio to ideal: 0.7694 -> 1.0112 on Walker2d | Helped |
| add hierarchical experience replay (ablation) | KTM-DRL · Benchmark B | performance ratio to ideal: 0.2215 -> 0.9745 on Ant | Helped |
| add offline transfer on Ant (ablation) | KTM-DRL · Benchmark B Ant | performance ratio to ideal: 0.9767 -> 0.9745 | No clear effect |
Closest papers by meaning
- A Transfer Approach Using Graph Neural Networks in Deep Reinforcement Learning (AAAI 2024 · similarity 0.86)
- Mastering Massive Multi-Task Reinforcement Learning via Mixture-of-Expert Decision Transformer (ICML 2025 · similarity 0.86)
- Continual Knowledge Adaptation for Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- SF-DQN: Provable Knowledge Transfer using Successor Feature for Deep Reinforcement Learning (ICML 2024 · similarity 0.85)
- Representation Learning Enables Scalable Multitask Deep Reinforcement Learning (NeurIPS 2026 · similarity 0.85)
- Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning (ICML 2025 · similarity 0.85)
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.