Research map / Reinforcement learning / Visual and representation RL
Galactic: Scaling End-to-End Reinforcement Learning for Rearrangement at 100k Steps-per-Second
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 |
|---|---|---|---|
| Train with Galactic rather than Habitat 2.0 | Rearrangement RL · Rearrange, eval episodes in Habitat 2.0 | success rate: 0.00 { }$ \pm$ 0.00 -> 26.40 { }$ \pm$ 0.43 | Helped |
| Train and evaluate the end-to-end policy in Galactic rather than Habitat 2.0 | End-to-end rearrangement policy · Rearrange, eval episodes | success rate: 0% -> 86.70 { }$ \pm$ 1.06 | Helped |
| Add safe drop and stop constraints (ablation) | Galactic rearrangement policy · Rearrange, 1000 evaluation episodes | success rate: 79.63% -> 84.73% | Helped |
| Remove distractor objects (ablation) | Galactic [Safe Drop/Stop] rearrangement · Rearrange | success rate: 84.73% -> 99.5% | Helped |
| Use a kinematic arm controller instead of the dynamics-based torque controller (ablation) | Galactic policy transferred to Habitat 2.0 · Rearrange, eval episodes in Habitat 2.0 | success rate: 26.4% -> 29.7% | No clear effect |
| Increase training experience to 5 billion steps in Galactic | End-to-end rearrangement policy · Rearrange | success rate: 0% -> 85% | Helped |
Closest papers by meaning
- Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning (ICLR 2025 · similarity 0.86)
- Kinetix: Investigating the Training of General Agents through Open-Ended Physics-Based Control Tasks (ICLR 2025 · similarity 0.86)
- ManiSkill-HAB: A Benchmark for Low-Level Manipulation in Home Rearrangement Tasks (ICLR 2025 · similarity 0.85)
- Data Scaling Laws in Imitation Learning for Robotic Manipulation (ICLR 2025 · similarity 0.85)
- Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating (NeurIPS 2026 · similarity 0.84)
- EnerVerse: Envisioning Embodied Future Space for Robotics Manipulation (NeurIPS 2025 · similarity 0.84)
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.