SCI-PRM: A Tool Aware Process Reward Model for Scientific Reasoning Verification
For scientific reasoning tasks in biology, chemistry, and physics, this work addresses the lack of verification in tool usage and reasoning, providing a method to enhance foundation models' performance.
The paper introduces Sci-PRM, a process reward model for scientific reasoning that verifies tool usage and factual consistency. It achieves significant improvements in test-time scaling and reinforcement learning, enabling models to surpass previous performance ceilings.
While Process Reward Models (PRMs) have achieved remarkable success in mathematical reasoning, their application in complex scientific domains-such as biology, chemistry, and physics remains largely unexplored. Scientific problems demand not only logical rigor but also factual consistency and the precise usage of domain-specific tools, areas where current models often suffer from hallucinations and lack of verification. In this paper, we first construct SCIPRM70K, a large-scale dataset featuring Chain-of-Tool trajectories that explicitly interleave reasoning with the execution of scientific tools. Building upon this, we train an efficient reward model called Sci-PRM to provide fine-grained supervision on tool selection, execution accuracy, and result interpretation at each step in one inference. Experiments demonstrate that Sci-PRM significantly enhances foundation models in two key aspects: (1) it enables effective test-time scaling via Best-of-N selection; and (2) when integrated into Reinforcement Learning, it serves as a dense reward signal that mitigates the critical issue of advantage disappearance, allowing the model to break through existing performance ceilings.