PRISM: Preserving VLA Skills with Skill-Mixture Routing
What happened
Jiachen Tao and other authors propose PRISM, which represents reusable skills as parameter-level primitives: each skill family is stored as an independent low-rank action expert adapter, and at inference time they are combined and invoked by a chunk-wise VLM-conditioned router. The authors state that during composite training all primitives remain frozen, so each skill's action expert adapter is preserved. On RoboCasa365, PRISM achieves a 56.43% atomic skill success rate and a 20.00% long-horizon composite task success rate; as a comparison, standard π0.5-LoRA two-stage fine-tuning achieves 17.86% and 17.50%. The authors also state that the method was validated on the AgileX PIPER real-robot platform.
AI 根据报道生成 · 2 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.ROPRISM: Retentive VLA adaptation via in-context skill mixing routing
PRISM represents reusable skills as parameter-level primitives, storing each skill family as a dedicated low-rank action-expert adapter, and combines and invokes them at inference via a block-wise VLM-conditioned router. On RoboCasa365, PRISM achieves a 56.43% atomic skill success rate and a 20.00% success rate on long-horizon composite tasks, compared with 17.86% and 17.50% for standard two-stage π0.5-LoRA fine-tuning.
Activity for this event
- Current
- 3
- Peak
- 10Nov 9 14:00
- Past 24 hours
- -50%
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.