🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Large language models can write informal proofs, translate them into formal ones, and achieve SoTA performance in proving competition-level maths problems! LM-generated informal proofs are sometimes more useful than the human ground truth 🤯 Preprint: arxiv.org/abs/2210.12283 🧵

Can large language models write prompts…for themselves? Yes, at a human-level (!) if they are given the ability to experiment and see what works. arxiv.org/abs/2211.01910 with Yongchao Zhou, Andrei Muresanu, Ziwen, Silviu Pitis, Harris Chan, and Jimmy Ba (1/7)

Strongly agree with the points of synthetic data + verification by formal tools. Check out our recent paper accepted to ICLR 2024 that makes a small step towards that direction! openreview.net/forum?id=V5tdi…

Tired of over-optimized generations that stray too far from the base distribution? We present SLCD: Supervised Learning based Controllable Diffusion, which (provably) solves the KL constrained reward maximization problem for diffusion through supervised learning! (1/n)

I’m presenting two papers on value-based RL for post-training & reasoning on Friday at AI for Math Workshop @ ICML 2025 at #ICML2025! 1️⃣ Q#: lays theoretical foundations for value-based RL for post-training LMs; 2️⃣ VGS: practical value-guided search scaled up for long CoT reasoning. 🧵👇

This captures something fundamental we're seeing in AI right now! The shift from just scaling pre-training to scaling test-time compute is huge. Our Q# + VGS work shows how value-based methods can guide models through the vast implicit graphs of reasoning possibilities.