Research map / Reinforcement learning / Visual and representation RL
Generalizing Skills with Semi-Supervised Reinforcement Learning
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 |
|---|---|---|---|
| add S3G training on unlabeled MDPs | RL policy · obstacle | success rate: 65% -> 79% | Helped |
| add S3G training on unlabeled MDPs | RL policy · 2-link reacher | success rate: 75% -> 98% | Helped |
| add S3G training on unlabeled MDPs | RL policy · 2-link reacher with vision | success rate: 69% -> 92% | Helped |
| add S3G training on unlabeled MDPs | RL policy · half-cheetah | success rate: 56% -> 79% | Helped |
| learn reward with inverse RL (S3G) instead of supervised reward regression (ablation) | S3G · obstacle | success rate: 29% -> 79% | Helped |
| learn reward with inverse RL (S3G) instead of supervised reward regression (ablation) | S3G · 2-link reacher | success rate: 60% -> 98% | Helped |
| learn reward with inverse RL (S3G) instead of supervised reward regression (ablation) | S3G · 2-link reacher with vision | success rate: 85% -> 92% | Helped |
| learn reward with inverse RL (S3G) instead of supervised reward regression (ablation) | S3G · half-cheetah | success rate: 73% -> 79% | Helped |
| S3G learned reward vs true reward oracle | S3G · obstacle | success rate: 36% -> 79% | Helped |
| S3G learned reward vs true reward oracle | S3G · 2-link reacher | success rate: 80% -> 98% | Helped |
| S3G vs oracle | S3G · 2-link reacher with vision | success rate: 100% -> 92% | Hurt |
| S3G vs oracle | S3G · half-cheetah | success rate: 86% -> 79% | Hurt |
Closest papers by meaning
- Foundation Policies with Hilbert Representations (ICML 2024 · similarity 0.89)
- BC-IRL: Learning Generalizable Reward Functions from Demonstrations (ICLR 2023 · similarity 0.89)
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.89)
- Reinforcement Learning Guided Semi-Supervised Learning (NeurIPS 2024 · similarity 0.88)
- Unsupervised Zero-Shot Reinforcement Learning via Functional Reward Encodings (ICML 2024 · similarity 0.88)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 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.