Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Jingzhi Gong

TL;DR

This work tackles the trustworthiness challenge in LM-based code optimization by integrating human feedback through reinforcement learning from human feedback (RLHF). It presents a research design, grounded in a Knowledge Transfer Partnership, to develop a reliable, human-centered optimization framework that leverages RLHF, agentic workflows, and both open-source and proprietary LMs. Key deliverables include datasets with human feedback, practical optimization tools, and interactive demos, along with dissemination through journals, conferences, and public channels. The study aims to produce auditable, robust optimization processes that align with developer needs and industry objectives, advancing cooperative and human-centric AI in software engineering.

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.

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

TL;DR

This work tackles the trustworthiness challenge in LM-based code optimization by integrating human feedback through reinforcement learning from human feedback (RLHF). It presents a research design, grounded in a Knowledge Transfer Partnership, to develop a reliable, human-centered optimization framework that leverages RLHF, agentic workflows, and both open-source and proprietary LMs. Key deliverables include datasets with human feedback, practical optimization tools, and interactive demos, along with dissemination through journals, conferences, and public channels. The study aims to produce auditable, robust optimization processes that align with developer needs and industry objectives, advancing cooperative and human-centric AI in software engineering.

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.

Paper Structure

This paper contains 19 sections, 1 figure.

Table of Contents

  1. Introduction
  2. Research Design
  3. Research Team
  4. Core researcher
  5. Advisors
  6. Stakeholders
  7. Research Topic
  8. Research Ethics
  9. Research Questions (RQs)
  10. Research Protocols
  11. Research Data
  12. Research Techniques
  13. Research Tools
  14. Research Impact
  15. Deliverables
  16. ...and 4 more sections

Figures (1)

  • Figure 1: Overview of the research design canvas.