BWhat changed
Built a math-specialized 7B language model using a selected large math corpus and introduced group relative policy optimization, which estimates a reward baseline from grouped outputs without a separate critic model.
Scientific scope
Built a math-specialized 7B language model using a selected large math corpus and introduced group relative policy optimization, which estimates a reward baseline from grouped outputs without a separate critic model. Scope limitation: The model and GRPO are treated as one paper-level contribution package. The paper's MATH and related benchmark results are self-reported and do not establish general mathematical reliability or proof correctness.
Evidence
The retained primary-source abstract, bibliographic record and stated review scope support this provisional assessment. No independent experimental replication by WRS is claimed. The model and GRPO are treated as one paper-level contribution package. The paper's MATH and related benchmark results are self-reported and do not establish general mathematical reliability or proof correctness.
Contributors
- The named authors listed in the linked research record · SCIENCE: Complete source byline retained in the research record; human research direction, method development and evaluation. Individual labor shares are not measured.
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- DeepSeekMath official code and model repository
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction