Training-free diffusion noise selection method SDN improves VLA policy robustness
What happened
Duc Minh Nguyen and other authors propose a training-free method, SDN (Selected Diffusion Noise), which treats diffusion noise as a controllable degree of freedom to improve the reliability of VLA policies at test time. SDN samples multiple candidate action chunks and performs two-stage selection: first, it filters using a counterfactual negative set according to task-relevant visual grounding, and then selects the kinematically most stable action chunk by JerkRMS. The authors report that in simulation and real-world experiments with three VLA policies, SDN achieved improvements of 8-10% and 18.33%, respectively.
AI 根据报道生成 · 6 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROSDN: Improving VLA policy robustness at test time via selected diffusion noise
The authors propose Selected Diffusion Noise (SDN), a training-free method that treats diffusion noise as a controllable degree of freedom to improve the test-time reliability of VLA policies. SDN samples multiple candidate action chunks and performs two-stage selection: it first filters by task-relevant visual grounding using a counterfactual negative set, then selects the kinematically most stable action chunk based on JerkRMS. In simulation and real-world experiments with three VLA policies, it achieved improvements of 8-10% and 18.33%, respectively.
Activity for this event
- Current
- 3
- Peak
- 10Nov 5 14:00
- Past 24 hours
- -49%
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.