Benchmark-Driven Selection of AI: Evidence from DeepSeek-R1

Petr Spelda, Vit Stritecky

Evaluation of reasoning language models gained importance after it was observed that they can combine their existing capabilities into novel traces of intermediate steps before tas…