Process Supervision-Guided Policy Optimization for Code Generation

TL;DR

This paper introduces a Process Reward Model (PRM) that provides dense, line-level feedback during code generation, significantly improving reinforcement learning-based code synthesis, especially for complex, long-horizon tasks.

Contribution

The paper presents a novel PRM that offers dense feedback during code generation, enhancing RL training efficiency and effectiveness over sparse reward methods.

Findings

PRMs improve RL code generation performance

Using PRMs as dense rewards boosts learning efficiency

PRMs are especially effective for long-horizon tasks

Abstract

Reinforcement learning (RL) with unit test feedback has enhanced large language models' (LLMs) code generation, but relies on sparse rewards provided only after complete code evaluation, limiting learning efficiency and incremental improvements. When generated code fails all unit tests, no learning signal is received, hindering progress on complex tasks. To address this, we propose a Process Reward Model (PRM) that delivers dense, line-level feedback on code correctness during generation, mimicking human code refinement and providing immediate guidance. We explore various strategies for training PRMs and integrating them into the RL framework, finding that using PRMs both as dense rewards and for value function initialization significantly boosts performance. Our experimental results also highlight the effectiveness of PRMs in enhancing RL-driven code generation, especially for long-horizon scenarios.