Artificial Intelligence in Law

Posts

• Reasoning Over Recall (12 Apr, 2026)
  A plain-English breakdown of why providing statutory text without accompanying case law degrades LLM ensemble performance, introducing the concept of "Statutory Myopia."
• Technical Setup of "Reasoning Over Recall" (12 Apr, 2026)

Talks

• Judgement Day: Benchmarking “Black Box” LLMs With Open Legal Datasets | Open Source Summit India, 2026
  Introduced "Statutory Myopia," demonstrating a 9 percentage point drop in accuracy when models over-rely on retrieved text. Benchmarked GPT-5, DeepSeek R1, and Gemini 2.5 Pro across 239 unique cases from the LegalBench dataset.
  View Presentation Slides | Watch the Talk

Papers & Preprints

• Reasoning Over Recall: Why Statute-Only RAG Degrades Performance on Doctrinal Legal Tasks (Preprint, 2026)
  DOI: 10.5281/zenodo.19522950

Projects

• LeginAI | Multi-LLM Reasoning System
  View Project