5 papers
ZeroCoder: Can LLMs Improve Code Generation Without Ground-Truth Supervision?
Lishui Fan, Mouxiang Chen, Tingwei Zhu +4
Code generation is important in software engineering, and Reinforcement Learning with Verifiable Rewards (RLVR) is a powerful paradigm to improve it through execution-based feedbac…
ReCode: Reinforcing Code Generation with Reasoning-Process Rewards
Lishui Fan, Yu Zhang, Mouxiang Chen +1
In practice, rigorous reasoning is often a key driver of correct code, while Reinforcement Learning (RL) for code generation often neglects optimizing reasoning quality. Bringing p…
Balancing Latency and Accuracy of Code Completion via Local-Cloud Model Cascading
Hanzhen Lu, Lishui Fan, Jiachi Chen +3
Line-level code completion requires a critical balance between high accuracy and low latency. Existing methods suffer from a trade-off: large language models (LLMs) provide high-qu…
FGIT: Fault-Guided Fine-Tuning for Code Generation
Lishui Fan, Zhongxin Liu, Haoye Wang +3
Modern instruction-tuned large language models (LLMs) have made remarkable progress in code generation. However, these LLMs fine-tuned with standard supervised fine-tuning (SFT) so…
Self-Explained Keywords Empower Large Language Models for Code Generation
Lishui Fan, Mouxiang Chen, Zhongxin Liu
Large language models (LLMs) have achieved impressive performance in code generation. However, due to the long-tail distribution of LLMs' training data, low-frequency terms are typ…