[2604.09482] Process Reward Agents for Steering Knowledge-Intensive Reasoning

Computer Science > Artificial Intelligence arXiv:2604.09482 (cs) [Submitted on 10 Apr 2026] Title:Process Reward Agents for Steering Knowledge-Intensive Reasoning Authors:Jiwoong Sohn, Tomasz Sternal, Kenneth Styppa, Torsten Hoefler, Michael Moor View a PDF of the paper titled Process Reward Agents for Steering Knowledge-Intensive Reasoning, by Jiwoong Sohn and 4 other authors View PDF HTML (experimental) Abstract:Reasoning in knowledge-intensive domains remains challenging as intermediate steps are often not locally verifiable: unlike math or code, evaluating step correctness may require synthesizing clues across large external knowledge sources. As a result, subtle errors can propagate through reasoning traces, potentially never to be detected. Prior work has proposed process reward models (PRMs), including retrieval-augmented variants, but these methods operate post hoc, scoring completed trajectories, which prevents their integration into dynamic inference procedures. Here, we introduce Process Reward Agents (PRA), a test-time method for providing domain-grounded, online, step-wise rewards to a frozen policy. In contrast to prior retrieval-augmented PRMs, PRA enables search-based decoding to rank and prune candidate trajectories at every generation step. Experiments on multiple medical reasoning benchmarks demonstrate that PRA consistently outperforms strong baselines, achieving 80.8% accuracy on MedQA with Qwen3-4B, a new state of the art at the 4B scale. Importantly,...