DOoM: Difficult Olympiads of Math
This provides a new benchmark for assessing language models in Russian STEM domains, but it is incremental as it adapts existing benchmarking approaches to a specific language and subject area.
The authors introduced DOoM, an open-source benchmark for evaluating language models on Russian mathematics and physics problems ranging from school to Olympiad levels, finding correlations between performance and token usage and differences between math and physics tasks.
This paper introduces DOoM, a new open-source benchmark designed to assess the capabilities of language models in solving mathematics and physics problems in Russian. The benchmark includes problems of varying difficulty, ranging from school-level tasks to university Olympiad and entrance exam questions. In this paper we discuss the motivation behind its creation, describe dataset's structure and evaluation methodology, and present initial results from testing various models. Analysis of the results shows a correlation between model performance and the number of tokens used, and highlights differences in performance between mathematics and physics tasks.