Yixiang Zhu, Ruilong Wu, Yonghao Chen, Jiaxiang Zou, Yusong Hu, Xinyu Chen · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28019 · open-access PDF
Vision-language-action (VLA) 정책은 일반적으로 실제 환경에서 비동기적 인 추론으로 배포되며, 다음 행동 조각이 로봇이 이전 조각을 실행하는 동안 생성됩니다. 새로운 조각이 헌신 될 때까지 로봇과 장면이 변경되므로 조각은 안정적인 관찰에 조건화되며 작업 성공은이 추론 지연이 증가함에 따라 급격히 떨어집니다. 다중 후보 조각을 생성하면 최종 선택을 알릴 수 있지만 기존의 다중 후보 방법은 안정적인 추론 시작 관찰을 가진 후보자를 순위로 평가하거나 반복적으로
Machine translation (Meta M2M100); the English original follows.
Vision-language-action (VLA) policies are commonly deployed with asynchronous inference in real-world settings, where the next action chunk is generated while the robot is still executing the previous one. By the time the new chunk is committed, the robot and scene have changed, so the chunk is conditioned on a stale observation and task success drops sharply as this inference delay grows. Generating multiple candidate chunks could let newer observations inform the final choice, but existing multi-candidate methods either rank candidates with the stale inference-start observation or repeatedly
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28019, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange