Event

LOGIV: A Task Graph Verification and Causal Local Repair Framework

1 reports · 1 sources

Enqiao Lu and co-authors released a paper proposing LOGIV, a framework for long-horizon robotic tasks: it uses an explicit task graph to connect symbolic planning, observation-based verification, and embodied execution, and during execution checks the physical validity of condition support and resource-ordering edges; after a failure, dependency tracing determines which parts need to be recovered. The framework runs on top of existing vision-language-action (VLA) and world action model (WAM) policies and adds no trainable components. The results reported by the authors are: on 20 dual-arm tasks in RoboTwin V2 and BiCoord, across three execution backbones, average task success rates improved by 10.67 and 15.33 percentage points respectively when no perturbations were injected; on three real-robot tasks, the number of successful trials increased from 12/30 to 17/30. These figures are the authors' own claims and await independent verification.