Event
Proposing Component Decorrelation Loss to Improve Compositional Diffusion Policies
1 reports · 1 sources
Alec Laprevotte and other researchers propose a plug-and-play auxiliary loss called score-decorrelation for multi-task diffusion policies that linearly combine scores, encouraging component specialization by penalizing overlap between component noise predictions, with almost no architectural changes and very low computational overhead. The researchers state that experiments on multiple benchmarks show this loss consistently reduces duplicate counting between components, improves task decomposition, and produces behaviorally meaningful specialization without affecting multi-task performance.
Report timeline
- Learning to Be Different: A Component Specialization Method for Compositional Diffusion Policies
A study proposes a plug-and-play auxiliary loss called score-decorrelation to address the problem that diffusion policies with linearly composed scores tend to have convergent component predictions that are hard to interpret in multitask robot learning. The loss encourages specialization by penalizing overlap between component noise predictions, with almost no architectural changes and very low computational overhead. Experiments on multiple benchmarks show that the method consistently reduces duplicate counting between components, improves task decomposition, and produces behaviorally meaningful specialization without compromising multitask performance.
arXiv cs.RO ·