TLDR AI · RESEARCH & INNOVATION

DEEPSEEK PROVER

arxiv.org27 min read

Summary from the TLDR AI newsletter

DeepSeek has produced an incredibly powerful math model that was trained via process supervision and dramatically outperforms larger models on various MATH benchmarks.