论文笔记(七十)DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning(二)
DeepSeek-R1:IncentivizingReasoningCapabilityinLLMsviaReinforcementLearning(二)文章概括摘要:2.方法2.3.DeepSeek-R1:冷启动强化学习2.3.1.冷启动2.3.2.面向推理的强化学习2.3.3.拒绝采样与监督微调2.3.4.面向所有场景的强化学习2.4.蒸馏:赋予小模型推理能力文章概括引用:@article{g