LLM adapted for Indian law achieves 60% on bar exam, beats GPT-3.5

Researchers have developed a framework called Legal Assist AI to address the gap in legal assistance access in India. This system utilizes a smaller, 8-billion-parameter quantized Llama 3.1 model, enhanced with a Retrieval-Augmented Generation (RAG) system and prompt engineering. The framework integrates over 600 legal documents, including recent legislation like the Bharatiya Nyaya Sanhita, and successfully mitigates hallucinations. It achieved a score of 60.08% on the All-India Bar Examination benchmark, outperforming GPT-3.5 Turbo, and demonstrated 22 times greater parameter efficiency. AI

影响 Demonstrates a cost-effective approach to building specialized legal AI tools, potentially improving access to justice.

排序理由 This is a research paper detailing a novel framework and benchmark results for domain-specific LLM adaptation.

在 arXiv cs.CL 阅读 →

• All-India Bar Examination
• Bharatiya Nagarik Suraksha Sanhita
• Bharatiya Nyaya Sanhita
• GPT-3.5 Turbo
• Jatin Gupta
• Legal Assist AI
• Llama 3.1

• 论文
• 产品
• 其他

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →