Summary of Athena: Retrieval-augmented Legal Judgment Prediction with Large Language Models, by Xiao Peng et al.
Athena: Retrieval-augmented Legal Judgment Prediction with Large Language Models
by Xiao Peng, Liang Chen
First submitted to arxiv on: 15 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Recently, large language models (LLLMs) have excelled in various domains, including legal scenarios. Prompt engineering (PE) has emerged as a crucial interface between LLMs and real-world applications. To overcome challenges like few-shot prompting, chain-of-thought, and retrieval-augmented generation (RAG), researchers have proposed various PE methods. However, RAG for legal judgment prediction (LJP) remains underexplored. This study proposes “Athena”, a novel framework combining RAG as a core preprocess component to enhance LLMs’ performance on specialized tasks. Athena constructs a knowledge base for accusations and employs a semantic retrieval mechanism through vectorization. Experimental results show that Athena’s overall performance has improved significantly, achieving state-of-the-art results on the CAIL2018 dataset. Ablation studies on the in-context window size parameter reproduce the “lost-in-the-middle” phenomenon with relative positional variations. With moderate hyperparameter tuning, accuracy can reach up to 95%. The study also investigates query rewriting and data distribution, providing potential directions for future research. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Recently, computer models have become very good at understanding human language. They’re used in many areas, including law. To help these models work better with real-world applications, experts have developed “prompt engineering” (PE). This involves finding the right way to ask a question or give instructions to the model. However, there’s still more to be done in this area, especially when it comes to using these models for making legal judgments. The researchers in this study created a new system called “Athena”. It uses a combination of techniques to help computer models make better predictions about legal cases. They tested Athena and found that it works very well, beating the previous best results on a specific dataset. The study also looked at how different parameters affect the accuracy of the model’s predictions. |
Keywords
» Artificial intelligence » Context window » Few shot » Hyperparameter » Knowledge base » Prompt » Prompting » Rag » Retrieval augmented generation » Vectorization
