PRDProductsStrategy
MAST
2 mentions across all digests
MAST (Multi-Agent System Failure Taxonomy) is a structured failure classification framework developed by IBM Research and UC Berkeley to categorize why enterprise AI agents fail, identifying 'Incorrect Verification' as the strongest predictor of failure across models.
/// Stats
First Seen2026-03-24
Last Seen2026-04-15
Total Mentions2
Subject Mentions2
Last 7 Days0
Sources2
Peak Relevance4/5
Active Predictions1
/// Recent Stories
2026-04-15HIGH
Britain's atomic brain trust gives itself till 2030 to unpick fusion challenges
UK commits £2.5 billion and a 2030 deadline to solve four critical fusion barriers—plasma performance, tritium-free fuel cycles, system integration, and commercial viability—via the STEP and MAST projects.
2026-03-21HIGH
IBM and UC Berkeley Diagnose Why Enterprise Agents Fail Using IT-Bench and MAST