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

TL;DR

This paper introduces EvalAct, a method that improves retrieval-augmented agents by explicitly evaluating each retrieval step and using these evaluations to better optimize multi-step reasoning, leading to higher accuracy.

Contribution

The paper proposes EvalAct, a novel framework that converts retrieval quality assessment into explicit actions and introduces PCAR for improved optimization in retrieval-augmented agents.

Findings

EvalAct achieves state-of-the-art accuracy on seven QA benchmarks.

Significant improvements are observed on multi-hop question answering tasks.

Explicit evaluation loops and PCAR contribute to the performance gains.

Abstract

Retrieval-augmented agents can query external evidence, yet their reliability in multi-step reasoning remains limited: noisy retrieval may derail multi-hop question answering, while outcome-only reinforcement learning provides credit signals that are too coarse to optimize intermediate steps. We propose \textsc{EvalAct} (Evaluate-as-Action), which converts implicit retrieval quality assessment into an explicit action and enforces a coupled Search-to-Evaluate protocol so that each retrieval is immediately followed by a structured evaluation score, yielding process signals aligned with the interaction trajectory. To leverage these signals, we introduce Process-Calibrated Advantage Rescaling (PCAR), a GRPO-based optimization method that rescales advantages at the segment level according to evaluation scores, emphasizing reliable segments while updating uncertain ones conservatively. Experiments on seven open-domain QA benchmarks show that \textsc{EvalAct} achieves the best average accuracy, with the largest gains on multi-hop tasks, and ablations verify that the explicit evaluation loop drives the primary improvements while PCAR provides consistent additional benefits.