search

FudanDISC / DISC-LawLLM

[中文法律大模型] DISC-LawLLM: an intelligent legal system powered by large language models (LLMs) to provide a wide range of legal services.
Apache License 2.0
549 stars 61 forks source link

Legal Element Extraction #30

Open mchen644 opened 11 months ago

mchen644 commented 11 months ago

Hello, very great job! Thank you for your contributions! Could you provide some more details on how to construct the legal element extraction dataset? If so, I would highly appreciate it! By the way, I want to find out if you have encountered the decrease of general ability while fine-tuning?

Charlie-XIAO commented 11 months ago

Could you provide some more details on how to construct the legal element extraction dataset?

There are multiple open-source datasets that you can refer to, for instance, LEVEN and JointExtraction. Based on these, we use ChatGPT to provide some chain-of-thoughts.

By the way, I want to find out if you have encountered the decrease of general ability while fine-tuning?

Yes, but I cannot summarize a method to avoid this from happening. As some suggestions, you may increase the diversity of your prompts, and mix some other types of data in each fine-tuning. Still, this is largely trial-and-error.