CoSineVerifier: Tool-Augmented Answer Verification for Computation-Oriented Scientific Questions

TL;DR

CoSineVerifier is a tool-augmented answer verifier that uses external computational tools to improve verification accuracy in scientific question answering, especially in algebra and physics domains, outperforming existing methods.

Contribution

The paper introduces CoSineVerifier, a novel tool-augmented verifier with a two-stage training pipeline that enhances verification in computation-oriented scientific questions.

Findings

Achieves state-of-the-art results on VerifyBench-Hard and SCI-Bench.

Outperforms existing verifiers in RLVR tasks on AIME'24 and AIME'25.

Demonstrates strong generalization across STEM and reasoning tasks.

Abstract

Answer verification methods are widely employed in language model training pipelines spanning data curation, evaluation, and reinforcement learning with verifiable rewards (RLVR). While prior work focus on developing unified verifiers applicable across multiple reasoning scenarios, significant challenges remain in computation-oriented scientific domains, such as algebraic equivalence checking and physical constant substitution. In this paper, we introduce \model, a tool-augmented verifier that leverages external executors to perform precise computations and symbolic simplifications. \model enables robust verification that goes beyond simple semantic matching. We propose a novel two-stage pipeline, which begin with cold-start fine-tuning and followed by multi-turn reinforcement learning with tool integration. Extensive experiments conducted on STEM subjects, general QA, and long-form reasoning tasks demonstrates strong generalization of \model. The results shows that the \model achieves state-of-the-art performance on VerifyBench-Hard and SCI-Bench. And we also employ our \model in RLVR as a reward model, the results show that it consistently outperforms both rubric-based and model-based verifiers on AIME'24 and AIME'25, demonstrating strong potential to enhance reasoning capabilities of LLM. Our model is released at \hyperlink{https://huggingface.co/Nanbeige/CoSineVerifier-Tool-4B}{https://huggingface.co/Nanbeige/CoSineVerifier-Tool-4B}.