Multi-agent autocompletion tasks where each model fills in part of a codebase. The LLM_Collab_Code_Completion project currently focuses on ClassEval, which asks teams of LLMs to finish class skeletons based on docstrings and partially implemented methods.