{"type":"ScholarlyArticle","title":"Fin-PRM: A Domain-Specialized Process Reward Model for Financial Reasoning in Large Language Models","authors":[{"name":"Jie Zhu","slug":"jie-zhu","given":"Jie","family":"Zhu","affiliation":""},{"name":"Yuanchen Zhou","slug":"yuanchen-zhou","given":"Yuanchen","family":"Zhou","affiliation":""},{"name":"Shuo Jiang","slug":"shuo-jiang","given":"Shuo","family":"Jiang","affiliation":""},{"name":"Junhui Li","slug":"junhui-li","given":"Junhui","family":"Li","affiliation":""},{"name":"Lifan Guo","slug":"lifan-guo","given":"Lifan","family":"Guo","affiliation":""},{"name":"Feng Chen","slug":"feng-chen","given":"Feng","family":"Chen","affiliation":""},{"name":"Chi Zhang","slug":"chi-zhang","given":"Chi","family":"Zhang","affiliation":""}],"abstract":"Process Reward Models (PRMs) supervise intermediate reasoning steps in large language models (LLMs), but existing PRMs are mainly trained on general-domain data and struggle with the structured, symbolic, and fact-sensitive nature of financial reasoning. Financial tasks require not only correct final answers but also verifiable intermediate steps grounded in domain knowledge. In this paper, we propose Fin-PRM, a domain-specialized, trajectory-aware PRM for financial reasoning that jointly models step-level correctness and trajectory-level coherence, producing binary supervision signals for both local and global reasoning quality. To support reliable supervision, we construct a high-quality financial reasoning dataset of 3K trajectories, where step- and trajectory-level labels are automatically derived from multi-source reward signals, including Monte Carlo rollouts, LLM-based evaluation, and explicit financial knowledge verification. Fin-PRM defines a unified ranking score that integrates step- and trajectory-level rewards, enabling consistent use across multiple settings. We evaluate Fin-PRM in three scenarios: (1) offline trajectory selection for supervised fine-tuning, (2) reward-guided Best-of-N inference for test-time scaling, and (3) process-aware reward shaping for reinforcement learning. Experiments on financial reasoning benchmarks, including CFLUE and FinQA, show that Fin-PRM consistently outperforms general-purpose PRMs and strong baselines. Our project resources will be available at https://github.com/aliyun/qwen-dianjin.","doi":"10.24963/ijcai.2026/683","url":"https://ijcai.org/proceedings/2026/683","pdf":"https://ijcai.org/proceedings/2026/0683.pdf","markdown":"https://ijcai.org/proceedings/2026/683.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/683","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"Natural Language Processing","track":"Main Track","number":683,"pages":{"first":6136,"last":6144},"keywords":["Natural Language Processing: Resources and evaluation","Natural Language Processing: Other","Natural Language Processing: Language models"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}