NeurIPS 2023 LLM Efficiency Fine-tuning Competition
This study addresses the problem of benchmark overfitting in LLM evaluation for researchers, but it is incremental as it confirms known issues without introducing new methods.
The analysis of the NeurIPS 2023 LLM fine-tuning competition found that top models overfit on benchmark datasets, highlighting limitations in current evaluation schemes for generative models and the importance of data curation for high performance.
Our analysis of the NeurIPS 2023 large language model (LLM) fine-tuning competition revealed the following trend: top-performing models exhibit significant overfitting on benchmark datasets, mirroring the broader issue of benchmark overfitting on popular leaderboards and that data curation is essential in order to get a high performing LLM. The competition, which consisted of two stages - an open evaluation stage with publicly available tasks and a closed evaluation stage with unseen tasks - allowed us to assess the generalizability of fine-tuned LLMs. Our results highlight the limitations of current benchmark-based evaluation schemes for generative models and demonstrate the need for more robust evaluation methods. Notably, the winning submissions utilized standard open-source libraries and focused primarily on data curation. To facilitate further research and promote reproducibility, we release all competition entries, Docker files, and evaluation infrastructure, providing a valuable resource for the community to explore fine-tuning, overfitting, and reproducibility in LLMs.