analiza aiqbench
This roundup is mostly about AI agent research: a self-improving framework (ForkPilot) for multi-step tasks, a benchmark generator (CABRA) for coding agents, and work on computer-use agents learning from activity logs. The reported efficiency gains and the finding that coding agents struggle more with understanding unfamiliar code than with modifying it are useful signals. Benchmark-specific results and cost claims should be checked against the underlying papers before being relied on.
Microsoft Research Focus – Edition #97
Host Sowbhagya Suresh shares some of the latest news and cutting-edge research from across Microsoft—from understanding the challenges facing coding agents to uncovering how people use AI at work and improving extended-range weather forecasting. The post Microsoft Research Focus – Edition #97 appeared first on Microsoft Research.
More and more, AI agents are handling tasks that are complex. Tasks that require multiple steps, decisions, and interactions with external tools. But how can we count on AI agents to do the work if their ability to adjust when things go wrong may require a lot of resources with no guarantees of a better result? A new self-improving framework called ForkPilot learns when it’s worth revisiting earlier decisions to find a different approach and adapts its strategy based on new outcomes. Tested across six benchmarks and leading AI models,…
Extras afișat conform termenilor RSS ai publicației · 90 cuvinte · textul integral și figurile sunt pe site-ul publicației
Subiecte conexe
One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact
Microsoft Research · 28 sept. 2026
Called to serve: Tech, research, and positive impact
Microsoft Research · 20 oct. 2026
Microsoft’s Skala AI Model Joins CP2K to Improve Quantum Mechanical Simulations
The Quantum Insider · 24 aug. 2026
What AI gets wrong and what failure teaches us
Microsoft Research · 6 oct. 2026