LLM4CODE

Software Heritage is participating in Inria’s LLM4CODE challenge, a strategic initiative aimed at building reliable and productive code assistants powered by Large Language Models (LLMs).

This challenge directly addresses the critical need for more robust AI tools in software engineering by focusing on three key work packages:

• Self-improving code generation: Pushing the boundaries of LLM capabilities to produce higher-quality, more reliable code automatically.
• Evolution of existing software: Developing AI tools that can safely and efficiently handle software migration, maintenance, and updates.
• Interactive tools with AI-in-the-loop: Creating human-AI interfaces that enhance developer confidence and integrate seamlessly with existing verification and testing tools.

Learn more on the project website: https://project.inria.fr/llm4code