研究人员推出了一种名为向量策略优化（VPO）的新型强化学习算法，旨在增强语言模型输出的多样性。与优化单一标量奖励的传统方法不同，VPO训练模型来预测和生成针对多个、向量值奖励函数定制的解决方案。这种方法旨在通过产生更多样的响应来提高复杂搜索过程中的性能，这对于代码生成和演进搜索策略等任务至关重要。 AI
影响 通过优化多样化的奖励函数，增强了LLM在复杂搜索任务中的适应性。
排序理由 该集群包含一篇详细介绍语言模型训练新算法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.22817v1 Announce Type: cross Abstract: Language models must now generalize out of the box to novel environments and work inside inference-scaling search procedures, such as AlphaEvolve, that select rollouts with a variety of task-specific reward functions. Unfortunatel…
Language models must now generalize out of the box to novel environments and work inside inference-scaling search procedures, such as AlphaEvolve, that select rollouts with a variety of task-specific reward functions. Unfortunately, the standard paradigm of LLM post-training opti…