#GPT-5

共 28 篇带此标签

28 articles Newest 2026-09-15
Archive · #GPT-5 11 articles
Opinion

Terry Tao Calls Out AI's Math Problem

LLMs ace competition math but stumble on real research. A Fields Medalist's critique exposes the gap between AI benchmarks and scientific usefulness.

2026-09-12 Source 2026-09-11 6 min 460 views
Frameworks

Terminal-Bench-Science: Can AI Agents Actually Run Experiments?

Terminal-Bench-Science pushes Agent evaluation beyond coding tasks into full scientific workflows, with 80+ real research tasks spanning biology, materials science, and ML. Early results show today's top models are far from ready to serve as research assistants.

2026-08-28 Source 2026-08-28 5 min 460 views