Research map / Reinforcement learning / Visual and representation RL
Coarse-To-Fine Q-Attention: Efficient Learning for Visual Robotic Manipulation via Discretisation
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 |
|---|---|---|---|
| Replace ARM's actor-critic next-best-pose agent with coarse-to-fine discrete Q-attention | ARM · RLBench, 8 front-camera tasks | task performance | Helped |
| Use a single front-facing camera instead of wrist and shoulder cameras (ablation) | C2F-ARM · RLBench, 4 multi-camera tasks | task performance | Hurt |
| Vary coarse-to-fine depth and voxel grid size (ablation) | C2F-ARM · Take lid off saucepan; put rubbish in bin | task performance | Mixed |
| Use the coarsest configuration, depth 2 with 8^3 voxel grids (ablation) | C2F-ARM · Take lid off saucepan; put rubbish in bin | task performance | Hurt |
| Increase voxel grid size (ablation) | C2F-ARM · Take lid off saucepan; put rubbish in bin | task performance | Helped |
| Increase coarse-to-fine depth rather than voxel grid size when memory is limiting (ablation) | C2F-ARM · Take lid off saucepan; put rubbish in bin | task performance | Helped |
Closest papers by meaning
- Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.85)
- Robot-R1: Reinforcement Learning for Enhanced Embodied Reasoning in Robotics (NeurIPS 2025 · similarity 0.85)
- ActQuant: Sub-4-bit Action-Guided Quantization for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.85)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.84)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (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.