English

DreamPRM-Code: Function-as-Step Process Reward Model with Label Correction for LLM Coding

Machine Learning 2025-12-18 v1 Artificial Intelligence Computation and Language

Abstract

Process Reward Models (PRMs) have become essential for improving Large Language Models (LLMs) via test-time scaling, yet their effectiveness in coding remains limited due to the lack of meaningful step decompositions in code and the noise of Monte-Carlo-generated partial labels. We propose DreamPRM-Code, a coding-focused PRM that treats functions as reasoning steps using a Chain-of-Function prompting strategy to induce modular code generation, enabling PRM training and application analogous to mathematical reasoning tasks. To address label noise, DreamPRM-Code introduces a meta-learning-based correction mechanism that leverages clean final-solution unit-test labels and performs bi-level optimization to refine intermediate labels. Applying on test-time scaling, DreamPRM-Code achieved state-of-the-art performance on LiveCodeBench with 80.9 pass@1 rate, surpassing OpenAI o4-mini.

Keywords

Cite

@article{arxiv.2512.15000,
  title  = {DreamPRM-Code: Function-as-Step Process Reward Model with Label Correction for LLM Coding},
  author = {Ruiyi Zhang and Peijia Qin and Qi Cao and Pengtao Xie},
  journal= {arXiv preprint arXiv:2512.15000},
  year   = {2025}
}
R2 v1 2026-07-01T08:28:24.427Z