4 ms·

How likely is it that Putnam answers were in DeepSeek's training data?

• EvgeniyZh 1y ago
  The solutions weren't published anywhere. There is also no good automatic way to generate solutions as far as I know, even expensive ones (previous sota was 10 solutions and one before was 8 using pass@3200 for 7b model). Potentially the developers could've paid some people who are good in putnam-level math problems and lean to write solutions for LLMs. It is hard to estimate likelihood of that but it sounds like waste of money given relatively marginal problem/benchmark.
  • HappyPanacea 1y ago
    AoPS seems to have a forum dedicated to Putnam (including 2024): https://artofproblemsolving.com/community/c3249_putnam https://artofproblemsolving.com/community/c3249_putnam and here is a pdf with solutions to Putnam 2023: https://kskedlaya.org/putnam-archive/2023s.pdf https://kskedlaya.org/putnam-archive/2023s.pdf
    • EvgeniyZh 1y ago
      These are still need to be formalized in Lean which can be harder than solving the problem sometimes