[ PAPER ] · 2025 · arXiv.org
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
Adam Suma, Sam Dauncey
No open-access PDF, so we'll narrate a briefing from the abstract instead.
[ TLDR ]
This work introduces first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1, which incorporates multi-stage training and cold-start data before RL and achieves performance comparable to OpenAI-o1-1217 on reasoning tasks.