Event
SCOPE Uses Failed and Successful Executions to Guide Program Editing and Repair
1 reports · 1 sources
Junyang Shu et al. propose SCOPE (Selective Certification of Program Edits), which uses failed and successful executions as complementary evidence to guide program revision: a frontier-guided program search annotates execution evidence on a shared diagnostic graph, distinguishing verified progress from unresolved checks; selective behavior certification uses stateful API replay to restore recorded responses and changes, checking whether edits preserve selected historically successful behaviors. The paper claims that SCOPE outperforms the strongest baseline on both simulated and real tasks, achieving average success rates of 80.66% and 84.48% on LIBERO-PRO and Robosuite, respectively, improvements of 6.51 and 3.48 percentage points over the baseline.
Report timeline
- SCOPE: A Selective Program Edit Certification Method for Embodied Heuristic Learning
SCOPE (Selective Certification Of Program Edits) uses failed and successful executions as complementary evidence to guide program revision: frontier-guided program search annotates a shared diagnostic graph with execution evidence to distinguish verified progress from unresolved checks, while selective behavioral certification uses stateful API replay to restore recorded responses and mutations and checks whether edits preserve selected historical successful behaviors.
arXiv cs.RO ·