Microsoft Discovery Engine with CLIO excels in adaptive AI for scientific discovery
Microsoft Discovery Engine, powered by CLIO (Cognitive Loop via In-Situ Optimization), achieved high scores on the Agent's Last Exam benchmark, evaluating long-running, tool-using professional tasks. This demonstrates the platform's capacity for adaptive reasoning, pursuing multiple hypotheses, and learning from evidence to accelerate complex scientific and engineering problems. The system scored 61.6% in health and medicine, 75.2% in physical sciences, and 64.6% in life sciences, providing a robust foundation for real R&D work. This innovation is available to R&D organizations across all industries and the scientific community, enhancing discovery processes without replacing human experts.
Notes (1) ›
- Microsoft Discovery Engine excels in scientific discovery benchmark
Microsoft Discovery Engine with CLIO (Cognitive Loop via In-Situ Optimization) achieved higher scores on Agent's Last Exam, a demanding benchmark for agentic AI. It scored 61.6% in health and medicine, 75.2% in physical sciences, and 64.6% in life sciences, showcasing its ability for adaptive reasoning in complex scientific problem-solving.
https://azure.microsoft.com/en-us/blog/beyond-the-benchmark-how-an-adaptive-approach-drives-scientific-discovery/
Related releases
- Public Preview: Reader Endpoint for Azure Database for MySQL Flexible Server Azure Updates ·
- Azure Playwright Workspaces Reaches GA in Australia East, Japan East, Switzerland North Azure Updates ·
- Generally Available: Azure Developer CLI (azd) Extension Framework Azure Updates ·
- Azure AI Foundry model MAI-Transcribe-1 2026-01-23 reaches end of life in 7 days endoflife.date ·
- Azure Functions runtime 1 reaches end of life in 7 days endoflife.date ·
- Microsoft Foundry optimizes enterprise AI agent costs via context engineering Microsoft Azure Blog ·