研究人员开发了一个新框架，旨在最大限度地减少大型语言模型（LLM）激活引导中的“附带损害”，以在不负面影响无关任务性能的情况下控制模型行为。另一篇论文介绍了一种模式感知累积过程奖励模型（SCPRM），通过更准确、风险更敏感地评估推理路径来改进知识图谱问答。此外，一种名为面向数据影响的树搜索（DITS）的新方法，通过识别对模型改进影响最大的数据来增强多智能体系统的训练，其性能优于仅依赖Q值的传统方法。 AI
影响 这些论文引入了改进LLM控制、知识图谱推理准确性和多智能体系统训练效率的新技术，有望带来更强大、更具能力的AI系统。
排序理由 该集群包含三篇在arXiv上发表的独立学术论文，重点关注LLM控制、知识图谱推理和多智能体系统训练方面的新研究。
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
arXiv:2605.02819v1 Announce Type: new Abstract: Large language models excel at complex reasoning, yet evaluating their intermediate steps remains challenging. Although process reward models provide step-wise supervision, they often suffer from a risk compensation effect, where in…
arXiv:2605.01167v1 Announce Type: new Abstract: Activation steering is a method for controlling Large Language Model (LLM) behavior by intervening in its internal representations to increase the alignment with a specific target feature direction. However, standard interventions, …
Large language models excel at complex reasoning, yet evaluating their intermediate steps remains challenging. Although process reward models provide step-wise supervision, they often suffer from a risk compensation effect, where incorrect steps are offset by later correct ones, …
Large language models excel at complex reasoning, yet evaluating their intermediate steps remains challenging. Although process reward models provide step-wise supervision, they often suffer from a risk compensation effect, where incorrect steps are offset by later correct ones, …
arXiv:2502.00955v2 Announce Type: replace Abstract: Monte Carlo Tree Search (MCTS) based methods provide promising approaches for generating synthetic data to enhance the self-training of Large Language Model (LLM) based multi-agent systems (MAS). These methods leverage Q-values …