RLTF: Reinforcement Learning from Unit Test Feedback

TL;DR

RLTF introduces an online reinforcement learning framework that leverages detailed unit test feedback to improve code generation models, achieving state-of-the-art results on major benchmarks.

Contribution

The paper presents RLTF, a novel online RL approach that uses multi-granularity unit test feedback for more effective code model training.

Findings

Achieves state-of-the-art performance on APPS benchmark.

Outperforms previous methods on MBPP benchmark.

Utilizes real-time data generation and fine-grained feedback.

Abstract

The goal of program synthesis, or code generation, is to generate executable code based on given descriptions. Recently, there has been an increasing number of studies employing reinforcement learning (RL) to improve the performance of large language models (LLMs) for code. However, current representative works either rely solely on offline frameworks, limiting the exploration of new sample spaces, or fall short in the utilization of unit test signals, not accounting for specific error locations within the code. To address these issues, we propose RLTF, i.e., Reinforcement Learning from Unit Test Feedback, a novel online RL framework with unit test feedback of multi-granularity for refining code LLMs. Our approach generates data in real-time during training and simultaneously utilizes fine-grained feedback signals to guide the model towards producing higher-quality code. Extensive experiments show that RLTF achieves state-of-the-art performance on the APPS and the MBPP benchmarks. Our code is available at: https://github.com/Zyq-scut/RLTF.