研究人员引入了 LamPO（Lambda Style Policy Optimization）和 LambdaPO，这是用于增强语言模型推理能力的新颖方法。这些方法通过使用成对分解优势，超越了传统的组相对目标，从而更好地捕捉响应质量的细微差别。在 Qwen3 和 Phi-4-mini 等模型上的各种基准测试实验表明，与现有方法相比，性能和训练稳定性均有所提高。 AI
影响 引入了更稳定、更高效的推理语言模型训练新技术。
排序理由 该集群包含两篇 arXiv 论文，详细介绍了改进语言模型推理能力的新方法。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
Reinforcement learning with verifiable rewards (RLVR) has become an effective paradigm for improving reasoning language models on tasks such as mathematics, coding, and scientific question answering. However, widely used group-relative objectives, such as GRPO, summarize each sam…
Group Relative Policy Optimization(GRPO) has become a cornerstone of modern reinforcement learning alignment, prized for its efficacy in foregoing an explicit value-critic by leveraging reward normalization across sampled trajectory cohorts. However, the method's reliance on a mo…