测试时搜索何时真正帮助世界动作模型？候选选择与模拟器完整性的实证审计

Authors:Jun Liu, Dajie Sun, Zhenglun Kong, Zhenjiang Fan, Peiyan Dong, Fan Yang, Wenbin Zhangg, Geng Yuan, Xue Lin, Yanzhi Wang, Dong Huang

View PDF HTML (experimental)

Abstract:World-action models increasingly apply test-time scaling, spending extra inference computation on candidate selection and adaptive allocation to improve control. Such gains are credible only when candidates carry exploitable signal, nominal compute savings become real speedup, and evaluation stays reliable. We audit all three on Cosmos Policy and RoboCasa with a premise-first, pre-registered methodology. On clean data, a learned value selector is 1.66 pp worse than random; a hindsight oracle gains only 1.40 pp, with an episode-level interval spanning zero; and 82% of states show identical candidate outcomes, leaving little selection headroom. Token pruning cuts per-forward FLOPs by up to 47% but yields at most a measured 1.15$\times$ end-to-end speedup on the deployed path (1.25$\times$ on a decode-excluded path), short of a 1.5$\times$ target, and our adaptive-allocation experiment is inconclusive under a power gate. We further identify and fix a silent rendering corruption whose rate rises sharply along compute-heavy planning paths while leaving the official direct-policy result unaffected, and release a detector, fix, and scripts. These findings show why test-time scaling should be audited as a chain of premises (decision signal, realized efficiency, evaluationintegrity) rather than by success rate alone. Conclusions are scoped to the tested configuration.

Subjects:	Artificial Intelligence (cs.AI)
Cite as:	arXiv:2610.41970 [cs.AI]
	(or arXiv:2610.41970v1 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2610.41970 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Jun Liu [view email]
[v1] Fri, 30 Oct 2026 16:21:07 UTC (1,568 KB)