coverage-aware

TRACE is a training-free framework that optimizes GUI agent efficiency by ranking visual evidence based on utility and diversity, reducing latency and memory usage through adaptive token management and KV contraction.

This paper introduces NVE, a separability and coverage-aware internal validation metric for biclustering, demonstrating its utility through synthetic and real datasets as a complementary criterion to existing coherence-based measures.

This paper proposes a coverage-aware active evaluation method for discovering failures in autonomous systems by combining proxy evaluations with limited target system testing, achieving up to 2x improvement in failure discovery over baselines.