Saltar al contenido
← volver

Agent Step Value: Auditing Evaluation-Channel Sensitivity in Black-Box Agent Traces

arXiv cs.AI · arxiv.org Publicado:

arXiv:2607.04419v5 Announce Type: replace Abstract: Reusing evaluator-derived step rewards across evaluation channels raises a measurement question: does the assigned direction remain stable? We define Agent Step Value (ASV) as a channel-indexed before/after target-margin gain and identify the state-by-channel interaction from a matched 2x2 design. Across frozen PubMed question-answering transitions, cross-channel gain signs disagree on 48.3% of comparisons, compared with 0.3% for same-channel…

Leer original →

Más de arXiv cs.AI