HackerPapers

arXiv:2402.03300cs.CL

DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y. K. Li, Y. Wu, Daya Guo

Abstract

Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. DeepSeekMath 7B has achieved an impressive score of 51.7% on the competition-level MATH benchmark without relying on external toolkits and voting techniques, approaching the performance level of Gemini-Ultra and GPT-4. Self-consistency over 64 samples from DeepSeekMath 7B achieves 60.9% on MATH. The mathematical reasoning capability of DeepSeekMath is attributed to two key factors: First, we harness the significant potential of publicly available web data through a meticulously engineered data selection pipeline. Second, we introduce Group Relative Policy Optimization (GRPO), a variant of Proximal Policy Optimization (PPO), that enhances mathematical reasoning abilities while concurrently optimizing the memory usage of PPO.

arXiv comments
Journal reference
DOI
Subjects
cs.CLComputation and Language cs.AIArtificial Intelligence cs.LGMachine Learning

On Hacker News

Linked by 7 identifiable commenters in 8 comments across 8 threads, from Feb 2024 to Sep 2026. Found in 6 HN submissions from Feb 2024 to Jan 2025. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments8 threads

  1. DeepSeek v4.1 Flash 1 comment
  2. Nanolang: A tiny experimental language designed to be targeted by coding LLMs 1 comment

    • Not exactly sure what you are looking for here. That GRPO works? > Group Relative Policy Optimization (GRPO), a variant reinforcement learning (RL) algorithm of Proximal Policy Optimization (PPO) (Schulman et al., 2017). GRPO foregoes the critic model, instead estimating the…

      nlView on HNParent context

  3. DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning [pdf] 1 comment

    • Well they do that too: https://huggingface.co/deepseek-ai/DeepSeek-Prover-V2-671B But I suppose the bigger goal remains improving their language model, and this was an experimentation born from that. These works are symbiotic; the original DeepSeekMath resulted in GRPO, which…

      natrysView on HNParent context

  4. DeepSeek’s founder is threatening US dominance in AI race 1 comment

    • You could maybe make that accusation about V3 (to the extent that it's a bad thing and not fair use, specially considering amoral origin of OpenAI's models in first place), but don't think the claim makes sense for R1 since OpenAI's o1 did not expose its CoT traces even in API.…

      natrysView on HNParent context

  5. Using GRPO to Beat o1, o3-mini and R1 at "Temporal Clue" 1 comment
  6. GPT-4.5 1 comment

    • It is possible to understand the mechanism once you drop the anthropomorphisms. Each token output by an LLM involves one pass through the next-word predictor neural network. Each pass is a fixed amount of computation. Complexity theory hints to us that the problems which are…

      porridgeraisinView on HNParent context

  7. Berkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Mod 1 comment

    • First graph tells the story - below a certain model size (500m params), reinforcement learning is close to useless. Above this (task-dependent) model size threshold, reinforcement learning basically works. I suspect this is what we saw play out with math/coding reasoning models…

      highfrequencyView on HNParent context

  8. DeepSeekMath 7B achieved 51.7% on MATH benchmark 1 comment

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.