Research map / Reinforcement learning / Visual and representation RL
Policy Contrastive Decoding for Robotic Foundation Models
Shihan Wu, Xu Luo, Ji Zhang, Junlin Xie, Jingkuan Song, Heng Tao Shen, Lianli Gao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 PCD with Grounding DINO annotation | OpenVLA · SIMPLER avg | success rate: 16.8 -> 25.3 (+50.6%) | Improved |
| add PCD with Grounding DINO annotation | Octo-base · SIMPLER avg | success rate: 13.8 -> 17.9 (+29.7%) | Improved |
| add PCD with Grounding DINO annotation | pi0 · SIMPLER avg | success rate: 63.9 -> 69.6 (+8.9%) | Improved |
| add PCD | pi0 fine-tuned · Real-world 6 tasks | success rate: +108% | Improved |
| add PCD | pi0 fine-tuned · Real-world 6 tasks | time cost: +24% | Hurt |
| add PCD | Octo-base · SIMPLER Google Robot Close Drawer | success rate: 31.0 -> 21.0 | Hurt |
| use Grounding DINO instead of Point/Box prompts for annotation (ablation) | PCD on OpenVLA/Octo · SIMPLER avg | success rate: OpenVLA 22.9/24.4 -> 25.3 | Improved |
| set contrast strength alpha > 0 (ablation) | PCD · SIMPLER avg | success rate | Improved |
| choose Grounding DINO over YOLO World or SED (ablation) | PCD · SIMPLER avg | success rate | Improved |
| use LaMa inpainting instead of Telea or Navier-Stokes (ablation) | PCD · SIMPLER avg | success rate | Improved |
Closest papers by meaning
- STRAP: Robot Sub-Trajectory Retrieval for Augmented Policy Learning (ICLR 2025 · similarity 0.88)
- Make-An-Agent: A Generalizable Policy Network Generator with Behavior-Prompted Diffusion (NeurIPS 2024 · similarity 0.87)
- Conditioning Matters: Training Diffusion Policies is Faster Than You Think (NeurIPS 2025 · similarity 0.86)
- DiffusionVLA: Scaling Robot Foundation Models via Unified Diffusion and Autoregression (ICML 2025 · similarity 0.86)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.86)
- Bidirectional Decoding: Improving Action Chunking via Guided Test-Time Sampling (ICLR 2025 · similarity 0.86)
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.
