3 comments

  • _ache_6 hours ago
    Already saw it on Mastodon yesterday. A better presentation is in the other link. (<a href="https:&#x2F;&#x2F;competition.sair.foundation&#x2F;competitions&#x2F;mathematics-distillation-challenge-equational-theories-stage1&#x2F;overview" rel="nofollow">https:&#x2F;&#x2F;competition.sair.foundation&#x2F;competitions&#x2F;mathematics...</a>)<p>I understand the idea of a distillation challenge but it really feels like the poor man&#x27;s solution to a problem that could be better solved by training a LLM and analysing the layers. In the end, as I value a condensed &quot;cheat sheet&quot;, if the goal is to improve open-source model. A better approach seems to recreate the AlphaProof system, longer to do, but more efficient. The path taken by mathematicians now is agentic system with general LLM.<p>- Masto: <a href="https:&#x2F;&#x2F;mathstodon.xyz&#x2F;@tao&#x2F;116225525978210807" rel="nofollow">https:&#x2F;&#x2F;mathstodon.xyz&#x2F;@tao&#x2F;116225525978210807</a><p>- AlphaProof: <a href="https:&#x2F;&#x2F;deepmind.google&#x2F;blog&#x2F;ai-solves-imo-problems-at-silver-medal-level&#x2F;" rel="nofollow">https:&#x2F;&#x2F;deepmind.google&#x2F;blog&#x2F;ai-solves-imo-problems-at-silve...</a><p>- AlphaProof description: <a href="https:&#x2F;&#x2F;www.nature.com&#x2F;articles&#x2F;s41586-025-09833-y" rel="nofollow">https:&#x2F;&#x2F;www.nature.com&#x2F;articles&#x2F;s41586-025-09833-y</a>
    • steppi5 hours ago
      It seems like the lift in the open-source models is being used as a proxy metric, and the core goal is a human understandable yoga [1] for approaching these kinds of equational proofs in universal algebra.<p>[1] <a href="https:&#x2F;&#x2F;mathoverflow.net&#x2F;questions&#x2F;64071&#x2F;what-does-the-term-yoga-mean-in-mathematics" rel="nofollow">https:&#x2F;&#x2F;mathoverflow.net&#x2F;questions&#x2F;64071&#x2F;what-does-the-term-...</a>
    • Copyrightest2 hours ago
      [dead]
  • sriramgonella4 hours ago
    [dead]
  • tug20246 hours ago
    [dead]