Evaluate-as-Action: Self-Evaluated Process Rewards for Retrieval-Augmented Agents

Researchers propose EvalAct, a new method that improves retrieval-augmented AI agents by converting retrieval quality assessment into explicit actions and using Process-Calibrated Advantage Rescaling (PCAR) for optimization. The approach shows superior performance on multi-step reasoning tasks across seven open-domain QA benchmarks by providing better process-level feedback signals.

• →EvalAct converts implicit retrieval quality assessment into explicit actions to improve multi-step reasoning in AI agents.
• →The method introduces a Search-to-Evaluate protocol where each retrieval is immediately followed by structured evaluation scores.
• →Process-Calibrated Advantage Rescaling (PCAR) optimizes advantages at segment level based on evaluation scores.
• →Testing on seven open-domain QA benchmarks shows EvalAct achieves best average accuracy with largest gains on multi-hop tasks.
• →The explicit evaluation loop drives primary improvements while PCAR provides consistent additional benefits.