ScholarPulse 日报 2026-10-16

2026-10-16 学术简报：2 篇。4个Gemini 4 Argon智能体通过自主编程与协作，在无人工干预下历时27小时14分击败末影龙，首度实现《我的世界》端到端自主通关。

今日速览

序号	标题	来源	日期	主题	推荐等级
1	A team of Gemini 4 Argon agents beats Minecraft	arXiv	2026-10-15	RAG	高
2	Learning to Steer LLM Agents and Multi-Agent Systems	arXiv	2026-10-15	AI-Agent	高

重点论文与技术动态

1. A team of Gemini 4 Argon agents beats Minecraft

来源：arXiv
日期：2026-10-15
作者/机构：Maxence Faldor, Satinder Singh, Roberta Raileanu
主题标签：RAG, arXiv
推荐等级：高
分类：cs.AI

一句话结论

4个Gemini 4 Argon智能体通过自主编程与协作，在无人工干预下历时27小时14分击败末影龙，首度实现《我的世界》端到端自主通关。

核心内容

完全自主协同与编程：无需预设角色或游戏逻辑，智能体靠游戏内聊天和共享内存分工协作，自主编写并调试了2,783个程序。
无干预完成通关：在简单难度的随机世界中历经605次死亡，仅依赖标准Mineflayer库击败末影龙，全程无人类介入或预选种子。
验证长程规划能力：成功维持跨越一天以上、含数千个依赖步骤的连贯行动，展现了契合现实复杂软件工程的长程编码与协作潜力。

方法与数据

4个智能体依托开源Mineflayer库控制角色，通过游戏内聊天与共享内存交互；测试环境为随机生成的简单难度Minecraft生存世界。

价值判断

值得关注：首个仅用通用标准库且无专用预制工具实现AI端到端通关Minecraft的成果。
可复用点：基于共享内存与聊天通信的多智能体自主规划、写码及调试范式。
局限/待核查：任务仅在简单难度下完成且死亡达605次；具体局限摘要未详述。

摘要

We placed four Gemini 4 Argon agents in a sandboxed Minecraft survival world and gave them one goal: defeat the Ender Dragon. Each agent controlled a player via an open-source library called Mineflayer and coordinated with its teammates through in-game chat and shared memory. No one assigned roles or wrote game logic. The agents planned, wrote, and debugged every program themselves. Researchers could watch the run on a live dashboard and pause or stop it at any time, but they had no cause to intervene. Starting from a randomly generated world on Easy difficulty, the team wrote 2,783 programs and defeated the dragon in 27 hours and 14 minutes, dying 605 times along the way. To our knowledge, this is the first end-to-end autonomous completion of Minecraft by AI agents, achieved using only the standard Mineflayer library without pre-scouted seeds or other game-specific tools written by humans. Winning required the agents to keep a coherent plan across thousands of dependent steps for more than a day. Complex real-world software engineering and enterprise tasks demand the same long-horizon coding and multi-agent coordination. We describe the setup, how the run unfolded, the team's strategies (dividing the work, sharing what they learned, recovering each other's items), and the limits of the result.

2. Learning to Steer LLM Agents and Multi-Agent Systems

来源：arXiv
日期：2026-10-15
作者/机构：Leyang Shen, Yaxin Gong, Yang Zhang, Chenxiao Fan, Jiakai Wang, Dong Wang
主题标签：AI-Agent, arXiv
推荐等级：高
分类：cs.LG, cs.AI, cs.MA

一句话结论

本文提出一种引导智能体（steering agent），通过快慢双路径机制在线纠正大模型智能体系统的“目标漂移”，显著提升长程任务的执行性能与效率。

核心内容

针对大模型单/多智能体在长程协作中易偏离任务目标的“目标漂移”痛点，提出实时监督与校正方案。
采用双路径决策结构：以低成本“快路径”判断何时介入，高成本“慢路径”生成具体干预策略。
学得的引导能力可泛化至分布外任务与未知执行器，且支持无需重新训练即可灵活调节干预频率。

方法与数据

联合强化学习训练：结合配对排序目标（pairwise ranking）与群体相对策略优化（GRPO）共同优化双路径策略。
实验设置：在信息检索类搜索任务（information-seeking search tasks）上评测了单智能体与多智能体系统。

价值判断

值得关注：构建了独立于执行器的通用外部干预范式，兼具强泛化性与频率可控性。
可复用点：快慢分流的干预决策架构，以及配对排序与 GRPO 结合的策略训练方法。
局限/待核查：实验评测仅明确提及信息检索搜索任务，对其他复杂应用场景的适用性摘要未明确说明。

摘要

Large language model (LLM)-based agents and multi-agent systems increasingly tackle complex tasks through long-horizon tool use and inter-agent collaboration. However, these systems are prone to deviating from the overall goal as execution unfolds, which wastes execution resources and leads to undesired outcomes, a failure termed goal drift. In this work, we propose a steering agent that supervises execution on the fly and steers the system back on track when drift occurs. The steering agent observes the execution context, deciding when to intervene with a low-cost fast path and how to intervene with a high-cost slow path. To unlock its steering intelligence, the ability to assess goal drift and generate effective interventions, we jointly train the two paths with reinforcement learning, combining a pairwise ranking objective with group-relative policy optimization (GRPO). Experiments on information-seeking search tasks show that the steering agent provides effective supervision for both single-agent and multi-agent systems, improving their performance and efficiency. The learned steering intelligence generalizes to out-of-domain tasks, transfers to unseen executors, and supports controllable intervention frequency without retraining.