These Mathematicians Are Trying to Educate A.I. - The New York Times

kept by eddie feb 08

Researchers are manually evaluating AI's failure to solve advanced math problems, revealing significant gaps in reasoning that automated benchmarks miss.

The AI Revolution in Math Has Arrived
Since mid-2025, AI models have begun proving new mathematical results at an accelerating pace, with mathematicians reporting discoveries that would have taken weeks now achieved in days.
also on #mathematics, #education, #ai-research, OpenAI
The end of the age of heroes - by Noah Smith - Noahpinion
AI’s recent breakthroughs in solving major math problems signal that human mathematical heroism is ending, reshaping the role of mathematicians.
also on #mathematics, #ai-research, OpenAI
Benchmark Scores = General Capability + Claudiness
Benchmark scores are dominated by a single 'general capability' dimension, with the second component mainly identifying a 'Claudiness' factor.
also on #benchmarks, #machine-learning, #ai-research, OpenAI
[2508.15260] Deep Think with Confidence
Deep Think with Confidence (DeepConf) uses model confidence signals to filter low-quality reasoning traces, achieving up to 99.9% accuracy on AIME 2025 with 84.7% fewer tokens.
also on #benchmarks, #machine-learning, #ai-research