Enhancing Trust in Language Model-Based Code Optimization through RLHF: A Research Design

TL;DR

This paper proposes a research design to improve trust in language model-based code optimization by integrating human feedback through reinforcement learning with human feedback (RLHF), addressing reliability issues like hallucinations.

Contribution

It introduces a novel research framework for enhancing trustworthiness of LMs in code optimization using RLHF, focusing on human-centric reliability improvements.

Findings

Framework for integrating human feedback into LM-based code optimization

Addressing hallucination issues in language models for software engineering

Lays groundwork for future empirical validation of trust-enhancing methods

Abstract

With the rapid advancement of AI, software engineering increasingly relies on AI-driven approaches, particularly language models (LMs), to enhance code performance. However, the trustworthiness and reliability of LMs remain significant challenges due to the potential for hallucinations - unreliable or incorrect responses. To fill this gap, this research aims to develop reliable, LM-powered methods for code optimization that effectively integrate human feedback. This work aligns with the broader objectives of advancing cooperative and human-centric aspects of software engineering, contributing to the development of trustworthy AI-driven solutions.