Aligning Crowd-sourced Human Feedback for Reinforcement Learning on Code Generation by Large Language Models
This work addresses the challenge of aligning AI tools like GitHub Copilot with human preferences for software developers, but it appears incremental as it builds on existing RLHF methods with a focus on crowd-sourcing and domain-specific applications.
The paper tackles the problem of improving text-to-code generation by large language models (LLM agents) through reinforcement learning with human feedback (RLHF), using a Bayesian optimization framework to align crowd-sourced human feedback. The result demonstrates effective training of LLM agents for enhanced code generation, though no concrete numbers are provided.
This paper studies how AI-assisted programming and large language models (LLM) improve software developers' ability via AI tools (LLM agents) like Github Copilot and Amazon CodeWhisperer, while integrating human feedback to enhance reinforcement learning (RLHF) with crowd-sourced computation to enhance text-to-code generation. Additionally, we demonstrate that our Bayesian optimization framework supports AI alignment in code generation by distributing the feedback collection burden, highlighting the value of collecting human feedback of good quality. Our empirical evaluations demonstrate the efficacy of this approach, showcasing how LLM agents can be effectively trained for improved text-to-code generation. Our Bayesian optimization framework can be designed for general domain-specific languages, promoting the alignment of large language model capabilities with human feedback in AI-assisted programming for code generation.