Research map / Reinforcement learning / Visual and representation RL
Unsupervised Reinforcement Learning of Transferable Meta-Skills for Embodied Navigation
Check 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 ULTRA's learned transferable meta-skills and curriculum-based training | Visual navigation policy · AI2-THOR test, All | SPL: 7.48 -> 11.47 | Helped |
| Add the shared hierarchical policy, separating task-specific master policy from shared sub-policies (ablation) | ULTRA · AI2-THOR test, All | SPL: 10.54 -> 11.47 | Helped |
| Add the Reptile meta-RL update to shared sub-policies (ablation) | ULTRA · AI2-THOR test, All | SPL: 11.03 -> 11.47 | Helped |
| Add curriculum-based adversarial task generation instead of random-location tasks (ablation) | ULTRA · AI2-THOR test, All | SPL: 8.35 -> 11.47 | Helped |
| Pretrain A3C on image-driven navigation using random target locations | A3C · AI2-THOR test, All | SPL: 7.48 -> 7.12 | Hurt |
| Pretrain A3C on image-driven navigation using a hand-crafted target curriculum | A3C · AI2-THOR test, All | SPL: 7.48 -> 8.04 | Helped |
Closest papers by meaning
- Unsupervised Curricula for Visual Meta-Reinforcement Learning (NeurIPS 2019 · similarity 0.91)
- HASARD: A Benchmark for Vision-Based Safe Reinforcement Learning in Embodied Agents (ICLR 2025 · similarity 0.89)
- METRA: Scalable Unsupervised RL with Metric-Aware Abstraction (ICLR 2024 · similarity 0.88)
- Grounding Video Models to Actions through Goal Conditioned Exploration (ICLR 2025 · similarity 0.88)
- PEAC: Unsupervised Pre-training for Cross-Embodiment Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Selective Visual Representations Improve Convergence and Generalization for Embodied AI (ICLR 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.