Nace AI's Drex for Search Tops Snowflake's MADQA Benchmark with 94.7% Accuracy on Enterprise Document Q&A
Sign in✓ Signed in
Nace AI's retrieval system for AI agents has reached the #1 position on Snowflake's MADQA benchmark, achieving 94.7% accuracy on a suite of 500 document-based question-answering tasks with Gemini 3.5 Flash as the reasoning model. The company's Drex for Search tool, designed to help autonomous agents locate exact answers in enterprise file collections, demonstrated a page-attribution F1 score of 88.2, indicating strong citation accuracy alongside answer correctness.
Continue reading
The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in✓ Signed in — this article isn’t included in your current plan.
Unlocking the full article...