Abstract
Most agent evaluations collapse a multi-step trace into a final answer, a success flag, or a trajectory-level score. This aggregation obscures the diagnostic question developers need most: which action changed the state in a useful direction? We introduce Agent Step Value (ASV), a state-transition measurement framework that scores each observed action by the change it induces in a state-grounded evaluator's distribution over fixed candidate outcomes.
ASV renders redacted before/after state projections, uses a stateless LLM evaluator to assign candidate log scores, and reports both gold-free belief diagnostics and offline oracle validation metrics. A label-free rationale pass separates evaluator deliberation from one-token option scoring, preserving candidate likelihoods while exposing leakage and floor-score events.
On 100 reviewed open-QA evidence-seeking tasks with live PubMed retrieval, a partially live DeepSeek actor, and DeepSeek log-probability scoring, ASV evaluates 1,100 steps and 2,200 states. Under the fixed-layout rationale-conditioned protocol, mean gold-margin gain is -2.335 (trajectory-bootstrap 95\% CI [-3.395, -1.272]), entropy movement is 0.000, and mean Bayesian surprise is 2.693. ASV therefore localizes constructive and destructive belief pivots that final-answer scores and entropy-only step metrics miss. We release the standalone ASV Eval toolkit.
Blogger's Review: The ASV framework offers a deeper perspective on agent evaluations by analyzing state transitions in detail. This method enhances understanding of agent behavior and effectively identifies key decision points, showcasing broad application potential, especially in complex decision systems. The release of its standalone toolkit will further propel research in this area.