DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

Notes

Nudged above paper prior (0.82) because the work was peer-reviewed and published in Nature (vol 645, pp 633-638, 2025) in addition to the arXiv preprint.

Evidence from this source (4)

• CDDeepSeek's R1 work demonstrates that LLM reasoning abilities can be incentivize… support
  weight

  0.85

  method: direct_measurement · locator: Abstract

  “the trained model achieves superior performance on verifiable tasks such as mathematics, coding competitions, and STEM fields, surpassing its counterparts trained via conventional supervised learning on human demonstrations.”

• CDUnder a pure-RL training regime on LLMs, advanced reasoning patterns including… support
  weight

  0.85

  method: direct_measurement · locator: Abstract

  “The proposed RL framework facilitates the emergent development of advanced reasoning patterns, such as self-reflection, verification, and dynamic strategy adaptation.”

• CDDeepSeek's R1 work demonstrates that LLM reasoning abilities can be incentivize… support
  weight

  0.90

  method: direct_measurement · locator: Abstract

  “Here we show that the reasoning abilities of LLMs can be incentivized through pure reinforcement learning (RL), obviating the need for human-labeled reasoning trajectories.”

• CDReasoning patterns that emerge in large RL-trained models can be systematically… support
  weight

  0.80

  method: direct_measurement · locator: Abstract

  “the emergent reasoning patterns exhibited by these large-scale models can be systematically harnessed to guide and enhance the reasoning capabilities of smaller models.”