SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2506.18896 · NeurIPS · 2025

ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs

Mengdi Wang, Jingrui He, Ling Yang, Jiahao Qiu, Jiaru Zou, Ke Shen, Jingwen Gu

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 5 functions out of this paper's own repositories and ran 0 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
yangling0818/buffer-of-thought-llm canonical 0 of 5
FunctionStatusWhere it lives
azure_openai_complete_if_cache Not yet run yangling0818/buffer-of-thought-llm/lightrag/llm.py
code served (permissive licence) · get_code("9768f858484cc024")
chunking_by_token_size Not yet run yangling0818/buffer-of-thought-llm/lightrag/operate.py
code served (permissive licence) · get_code("d91c95543a58a3c6")
extract_and_execute_code Not yet run yangling0818/buffer-of-thought-llm/meta_buffer_utilis.py
code served (permissive licence) · get_code("15dd38c0c4a5e24b")
extract_answer Not yet run yangling0818/buffer-of-thought-llm/meta_buffer_utilis.py
code served (permissive licence) · get_code("92d7b41138b8da5b")
openai_complete_if_cache Not yet run yangling0818/buffer-of-thought-llm/lightrag/llm.py
code served (permissive licence) · get_code("7e668b3293f54805")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Process Reward Models (PRMs) have recently emerged as a powerful framework for supervising intermediate reasoning steps in large language models (LLMs). Previous PRMs are primarily trained on model final output responses and struggle to evaluate intermediate thinking trajectories robustly, especially in the emerging setting of trajectory-response outputs generated by frontier reasoning models like Deepseek-R1. In this work, we introduce ReasonFlux-PRM, a novel trajectoryaware PRM explicitly designed to evaluate the trajectory-response type of reasoning traces. ReasonFlux-PRM incorporates both step-level and trajectory-level supervision, enabling finegrained reward assignment aligned with structured chain-of-thought data. We adapt ReasonFlux-PRM to support reward supervision under both offline and online settings, including (i) selecting high-quality model distillation data for downstream supervised fine-tuning of smaller models, (ii) providing dense process-level rewards for policy optimization during reinforcement learning, and (iii) enabling reward-guided Best-of-N test-time scaling. Empirical results on challenging downstream benchmarks such as AIME, MATH500, and GPQA-Diamond demonstrate that ReasonFlux-PRM-7B selects higher quality data than strong PRMs (e.g., Qwen2.5-Math-PRM-72B) and humancurated baselines. Furthermore, our derived ReasonFlux-PRM-7B yields consistent performance improvements, achieving average gains of 12.1% in supervised fine-tuning, 4.5% in reinforcement learning, and 6.3% in test-time scaling. We also release our efficient ReasonFlux-PRM-1.5B for resource-constrained applications and edge deployment.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2506.18896")
get_code_for_paper("2506.18896")
have("2506.18896")

Connect an agent — have() is free.