Users congratulated researcher Patty on the new benchmark testing AI agents for detecting hallucinated legal citations.
Based on 1 visible X reactions from 3 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@pattpatt_liu Congratulations Patty!
🚨 New Paper 🚀COLM! 1⃣ New benchmark for legal hallucination detection (even Claude Code doesn't catch them all)! 2⃣ Surprisingly, hallucination rates for free-tier ChatGPT versions through GPT-5 appear to *not* have improved, even as court cases with hallucinations skyrocket. https://twitter.com/pattpatt_liu/status/2075273821976731899
Claude Code and free-tier ChatGPT struggled to catch the hallucinations
Users congratulated researcher Patty on the new benchmark testing AI agents for detecting hallucinated legal citations.
Based on 1 visible X reactions from 3 accounts; directional sample.
Ask a question below.
Published answers will appear here.