Happy to introduce Kimina-Prover-72B ! Reaching 92.2% on miniF2F using Test time RL. It can solve IMO problems using more than 500 lines of Lean 4 code !
Check our blog post here:
https://t.co/QbrmoyYL9i
And play with our demo !
https://t.co/u0Wj0Id4vZ
We believe formal math is the future.
🔥Introducing Kimina-Prover Preview, a Numina &
@Kimi_Moonshot collaboration, the first large formal reasoning model for Lean 4, achieving 80.78% miniF2F.
https://t.co/fNX7orQYeZ
🔬 Sharing an early look at Kimina-Prover, our new Lean theorem proving model from our collaboration with Numina! @JiaLi52524397
🏆 Using an RL pipeline for proof exploration, Kimina-Prover Preview achieved 80.7% on the miniF2F — currently SOTA on this benchmark. We see promise in this approach for intuitive proof discovery.
Open-sourcing:
✅1.5B/7B distilled models
✅Custom autoformalizer
✅Revised miniF2F benchmark
Supporting the Lean community while we continue development. This is ongoing work, stay tuned for updates!
👉🏻 Full technical report on GitHub: 🔗 https://t.co/UZN29xJhKf
🚀 NuminaMath 1.5 is here! 🚀
900k+ high-quality competition math problems with CoT solutions, new problem metadata, manually verified Olympiad problems, and more! 📚🏅
Check it out: 🔗 https://t.co/2BQP7qdxJ7
Thanks to @Will424408@dsleo
Project Numina is thrilled to announce a €3m research grant from XTX Markets to support the development of open-source AI tools for mathematicians and general progress in AI reasoning!
https://t.co/wtm1skkFVu
🔴[ THREAD ] Les Jeux Olympiques devraient faire un retour aux sources et s’y tenir pour de bon.
▶️Oui c'est mon avis.
▶️Oui il y'a des paramètres à préciser.
▶️Oui c'est injuste (non.)
Mais comme on ne peut pas faire pire qu'aujourd'hui, je me permets ces hypothèses : ⬇️