AI
Microsoft Study Highlights Reliability Gaps in Long-Horizon AI Agents
A Microsoft Research study using the DELEGATE-52 benchmark reveals that frontier AI models lose up to 50% of document fidelity in extended workflows.
September 8, 2026, 02:50 AM·Crypto Briefing
