Tag
SOPO（Socratic Guided Policy Optimization）提出了一种多轮强化学习框架，通过苏格拉底式教师引导反馈和基于轮次的优势缩放来解决span级幻觉检测中的奖励稀疏与长度膨胀问题，在RAGTruth上取得SOTA，超越GPT-5等模型并开源了1.7B和4B模型。
EAU 是一个元智能体框架，在训练多轮 LLM 智能体时协同演化用户模拟器、环境与强化学习配置，使训练系统随策略一同进化。在三个交互式智能体基准上，EAU 比最强基线整体提升 14.7–15.0%，在 ColBench 上任务完成率最高提升 36.7%。
PlanPO is a reinforcement learning method that introduces coarse-to-fine advantage signals for multi-turn agentic LLMs, improving performance over GRPO by 27.2% on benchmarks like ALFWorld, WebShop, and SciWorld.
Proposes PaTR, a process-reward-guided adaptive tree rollout framework for multi-turn reinforcement learning in LLM agents. It selectively branches from promising states and prunes dead-end paths, achieving up to +5.0 on SWE-Bench and +9.3 on FrozenLake under the same training budget.
OpenWebRL presents an open framework for training visual web agents using online multi-turn reinforcement learning on real websites, achieving state-of-the-art performance with minimal initial supervision. Their 4B-parameter model outperforms prior open agents and competes with proprietary systems like OpenAI CUA and Gemini CUA.
A developer shares that addressing the 'tito' problem in multi-turn reinforcement learning is simpler than commonly believed, requiring only one implementation rule and a chat-template property that models already support.
DeepDive presents an automated approach to training deep search agents using knowledge graphs for data synthesis and multi-turn reinforcement learning, enabling complex multi-step reasoning and web browsing.
UI-TARS-2 is a native GUI-centered agent model that addresses data scalability, multi-turn RL, and environment stability challenges, achieving state-of-the-art results on GUI benchmarks (88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena,73.3 on AndroidWorld) and outperforming Claude and OpenAI agents.