Search papers, labs, and topics across Lattice.
This paper investigates the failure modes of action-chunked generative visuomotor policies, identifying "chunk-boundary artifact" as a key noise-sensitive failure mechanism. The authors demonstrate that artifact magnitude, which arises from discontinuities at chunk boundaries, is strongly correlated with task failure and can be systematically modulated by altering latent noise. By steering trajectories along artifact-related directions in noise space, they reliably alter artifact magnitude and, consequently, task success in hard-task settings.
Action-chunked policies can be steered towards success or failure by manipulating latent noise to control "chunk-boundary artifacts," revealing a surprising handle on these policies' reliability.
Action chunking has become a central design choice for generative visuomotor policies, yet the execution discontinuities that arise at chunk boundaries remain poorly understood. In a frozen pretrained action-chunked policy, we identify chunk-boundary artifact as a noise-sensitive failure mechanism. First, artifact is strongly associated with task failure (p<1e-4, permutation test) and emerges during the rollout rather than only as a post-hoc symptom. Second, under a fixed observation context, changing only latent noise systematically modulates artifact magnitude. Third, by identifying artifact-related directions in noise space and applying trajectory-level steering, we reliably alter artifact magnitude across all evaluated tasks. In hard-task settings with remaining outcome headroom, the success/failure distribution shifts accordingly; on near-ceiling tasks, positive gains are compressed by policy saturation, while the negative causal effect remains visible. Overall, we recast boundary discontinuity from an unavoidable execution nuisance into an analyzable, noise-dominated, and intervenable failure mechanism.
