Research Topic

I want to find recent papers on using generative AI for creating agent-like pipelines, especially in the context of coding assistants that can manage an entire codebase, including the architectural design, training methodologies, real-world applications, and performance metrics.

Summary

Several papers discuss generative AI for creating agent-like pipelines that manage entire codebases, including key aspects like architecture design, training methodologies, real-world applications, and performance metrics.

For instance, AutoCoder [1] excels in code evaluation through an agent interaction and execution-verified tuning approach, showing notable performance on the HumanEval benchmark. CodeAgent [3] details an LLM-based framework specifically for repository-level code generation with improved performance on real-world tasks. AutoDev [5] presents a comprehensive AI framework for automated software development, covering autonomous codebase management tasks. These papers collectively highlight the progression in employing generative AI to advance coding assistants capable of managing complete software development processes.

To understand the relationships and patterns within the papers found, see also:
So far, I've closely analyzed 540 of the most promising papers, and I've found ~32-84 that are relevant, which is probably ~82.4% of all that exist.
To get this estimate, we do a statistical analysis of the discovery process.

References