Reproduce “LLM-AutoDP: Automatic Data Processing via LLM Agents for Model Fine-tuning” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Wei Huang, Anda Cheng, Yinggui Wang, Lei Wang, Tao Wei

The authors introduce LLM-AutoDP, which uses an LLM agent to propose ordered data-processing pipelines for fine-tuning, evaluate them through downstream model performance, and revise them using comparative feedback. On Chinese-medical-dialogue, its processed-data Qwen2.5-7B-Instruct model beat the same-model No-Process baseline on 89.37% of GPT-4-judged question comparisons. This is a preference rate, not accuracy; results are limited to evaluated medical QA settings.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.