DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning (DeepSeekAI, 2025)

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• DeepSeek-R1-Zero spontaneously increased thinking time for difficult prompts, showing rudimentary meta-awarenessfinding0.821
  External finding cited as early demonstration of emergent self-regulatory potential resembling mindful self-monitoring
• Deep Reinforcement Learningmethod0.800
  AI training method inspired by behaviorism, used for autonomous cars and drones; cited as bioinspired success
• DeepSeek-R1framework0.782
  Open-source reasoning LLM from DeepSeekAI trained with reinforcement learning to exhibit self-reflection
• LLM judge (deepseek-v3) agrees with human evaluator on 91.6% of 200 sampled jailbreak responsesfinding0.770
  Validates the LLM-based harm evaluation rubric
• Reinforcement learning is sufficient for agency.claim0.769
  Argument that RL meets the agency indicator.
• Reinforcement Learning from Human Feedback (RLHF)framework0.760
  A competing alignment approach that fine-tunes models based on human evaluator feedback; discussed as complementary to SOO
• Reinforcement learning can be regarded as a limiting or special case of model-based approaches in general — or active inference in particular — when epistemic value is removed.claim0.757
  §3 Discussion.
• DeepSeek-R1 Llama 8b gains 0.16% accuracy on GSM8k with positive intervention (more reflections) at cost of ~2000 additional tokensfinding0.755
  Only model showing marginal benefit from increased reflection, at substantial token cost