Reproduce “Repository-Level Solidity Code Generation with Large Language Models: From Prompting to 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
Shi Chen, Rongcun Wang, Yuan Tian, Xiaoyuan Xie, Wei Song, Rubing Huang
The authors evaluate methods for generating complete Solidity contracts from natural-language requirements. They compare prompting, retrieval-augmented generation, and supervised fine-tuning, scoring outputs with BLEU and SolidityScore. For Qwen2.5-Coder-7B-Instruct, fine-tuning scores 35.9584 BLEU and 0.6465 SolidityScore, exceeding the best retrieval scores. These averages measure similarity, not security or deployability; only 31.3% of its fine-tuned outputs compile.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.