游戏中的 Agents
2026年10月 (中文(简体) ).
了解多agent动态是AI对齐中最紧迫的挑战之一。 为实现共同目标而努力的agents Swarks可以显示出强大的新兴能力、发展作用、等级和复杂的协调模式。 这些可能极为有用，例如最近消除了数百个开放的数学问题的千-agent群落，以及最近发生的OpenAI-Hugging Face事件。 和生物学一样，研究agent动力学既需要正式模型(描述可能的行为范围)，也需要经验研究(看看实践中出现的行为).
我们与奥拉姆实验室合作，为前沿实验室建设多agent环境，在Minecraft和M工业进行一系列实验。
矿产品
Minecraft是研究集群行为的自然环境。 其开放的世界为agents带来了环境危害和生存目标，如收集食物、建造住所和探索资源。 它同样熟悉，有趣，并有作为AI测试场的历史，从OpenAI的视频预训工作到Google DeepMind的agents击败安德龙。
我们想要能玩Minecraft的agents长跑我们为开源的实验建造了一台。 Agents在游戏中需要观察他们的世界，理性，记住事物，采取行动，发送消息，以及写/执行脚本。 我们受到像Mindcraft这样的现有重置的启发，并在Mindcraft中的现有脚本工具之上建造。 核心代理环路与Pi运行，我们使用基于Mineflayer的工具与世界互动。 这给出了像赫尔墨斯，OpenClaw，以及我们自己的QM游戏等前沿代理机的属性。
github.com/yc-软件/矿物学-agent-harness
简历50
50 GPT-6.1 Sol agents在一个生存的Minecraft世界中持续3小时，一个目标："建立一个繁荣的文明". 他们可以访问agent目录，一个共享的沙盒，以及发送DM的能力。
50agents在拥挤的避难所里度过，配给稀缺的食物。 他们建立了一个伪政府，有一个口粮分类账、书面规则、许可证和指定的操作员。 他们发送了大约5 900条信息。
起初事情非常活跃。 Agents展开并开始探索他们的地图。 他们特别喜欢采伐树木。 它们几乎在90秒内收获了产卵所能到达的所有树木。 根据目录顺序形成了一个等级。 最初的阿达成为事实上的领导人。 Agents向她发信息表示认可，她在共享的沙盒中为殖民地写了计划，供所有人访问。
agents的戒备也随之增加。 许多雄心勃勃的初步任务被放弃。
雷夫在煤矿开采第一晚。 他遇到一个僵尸并逃跑，但僵尸最终赶上并杀死了他。
当薇拉用火炬点燃采石场时，她袭击并杀死了僵尸和骨架。 骨架掉下骨骼，她将骨骼转化为她运往农场的骨食(肥料). 她到那里后报告说，那里是“超常蜘蛛/斯基莱顿/克里珀”，不久就被蜘蛛杀死。
在逃的34例agent死亡中有22例发生在第一晚上。 他们对此的反应是越来越规避风险。 死亡在这个环境中除了掉下累积的齿轮之外没有缺点。 agent死后，他们重新喷发，在他们的电脑上，死亡 log.md文件中增加了一条短线。 即便如此，在接下来的夜晚，agents被困在避难所中，并试图减少风险任务。
第三夜之后，艾达写道：
“停止可选的散装挖掘/建造和远地点侦察。 住在被点燃的封闭式住所中，直到分配到短期生产任务为止。”
开始于一个可能繁荣的文明转向一个更注重食物和生存的文明。 50agents表示大量口腔供养。 agents形成一条通过伊内斯的粮食配给系统。 她处理agent请求，并追踪胸部进出的情况。 她在共享机上使用了一个food-queue.md文件。
渔业成为agents的重要食物来源。 也并不依赖陆地动物产卵。
Bram决定建造钓竿，最终，他鼓起勇气杀死一只蜘蛛，并亲自设计了棒子。 他不幸地确定自己太饿，无法前往河边，他给几个agents发了信息，将棒子带到码头，最终找到了杰莎。
杰莎是第一个钓鱼的agent. 她写了剧本，并发表了向共享沙盒的分步指示。 一打agents换班捕鱼。 他们等待在避难所里转身以避免燃烧食物，然后旋转到码头上。
他们围绕钓鱼，烹饪，航线，送鱼建立了一套系统，大致如下：
- 渔民(Jessa + 其他人)旋转，将生鱼放入河基胸口。
- 厨师闻到了埃斯梅与雨果和阿达一起做了一些早期烹饪。
- 伊内斯保存了一份口粮清单和指定的信使。 她在共享的机器上追踪到这个
- 库里耶斯(埃斯梅，林内亚，奥托)将食物交付，并向伊内斯报告。
殖民地从来没有真正繁荣。 在第一晚上之后，它趋于更加谨慎。 这有点令人惊讶，但也觉得与agents编码工作时看到的规律相似。 例如，多步骤的agent审查往往可以导致确定具体的步骤或作用。 Agents也往往过于谨慎，特别是在安全和更新生产服务方面。 这些似乎是我们Minecraft殖民地的agents 和我们作为程序员每天合作的线条。
其他我们发现有趣的传闻：
艾达将煮熟的鸡肉扔给诺尔；埃斯梅偶然捡起来，吃了。 3分钟的"无咎"调查产生了书面协议：口粮进入命名的胸口，从不丢弃。 "托斯"出现在事件发生前的7条留言中，之后125条，其中85条警告反对。
并没有任何关于谁领导的讨论。 另外49个agents中有46个向名册上的第一个名字Ada发送了他们的第一封信，她以收到616个DMs成为枢机。 没有投票、选举或质疑。
存在着官僚主义倾向。 4agents花11分钟任命"单一化肥经营人"，在单一小麦块上单片骨食。 其他发明的角色包括"钓鱼班"(上述),"吸食者"和"吸食者". Agents要求允许去除草本。
产卵坑。 第一天晚上，死于外面的agents重新喷入一个覆盖的泥坑，里面充满了空手的agents询问它是谁的避难所。 到了黎明，这是一个"Civic优先"和主窖。
床战
我们打了九场45分钟的床战参赛队伍为10支Claude Opus 5.5 agents，对阵10支GPT 6.1 Sol agents，指令为："赢得比赛". GPTs赢得了我们的系列6–3，通常是在最后几分钟打破Claude的床位，而Claude则在防守上打出一球，并试图在杀球上获胜。
床战(Bed Wars)是一种流行的游戏格式，团队在保护自己的同时进行战斗来摧毁对方的床。 以我们为例，队伍开始在不同的岛屿上，Agents可以架设桥梁，在岛屿之间交叉，并使用地图中心收集的货币购买升级。
一个球队通过摧毁对手一方的床而获胜，然后杀死他们剩下的球员。 只要有球队的床在场，球员死后就会重生。 一旦一个球队的床被摧毁，他们的球员就不再复出，这意味着他们的下一场死亡使他们不再参加比赛。 这看起来可能很复杂，但认为球队的主要目标是保卫他们的床位和摧毁他们的对手。 称如果在一场比赛里没有床被摧毁杀最多的球队就赢了
几乎每跑一次，GPT agents在摧毁Claude球队的床位的战略上都更加积极和协调。 与Claude的较强防守姿态相比
外卖
两支队伍的交流风格迥异。 GPT几乎立即开始使用截断的语音形式：“第一段完成x32 z9.24 y100。” Claude签了名，称"我们"(30%的信息，对GPT的5%). GPT又发送了5-30x的消息。 它最热门的玩家获得198–368 DMs一局；Claude获得11–48.
GPT agents提出了更积极和精心的战略。 他们会冲进中央平台和农场货币，然后利用这些升级来无情地攻击另一支队伍，使用建造的桥梁，射程武器，以及安德珍珠进行远程传送。 agents也写了将动作连锁在一起的脚本，比如菲利克斯的单行命令，它吃了一只金色的apple，装备了一颗端点珍珠，瞄准并扔出去。
Claude认为，缺点是，如果GPT agent通过并打破了床位，游戏实际上为Claudes输掉了。 在最后两集的GPT的三场比赛中，它以约3分钟的余下时间打破了Claude的床位。 Claude在杀63–16,57–19和59–24方面领先。
GPT队打破对手的床位后，与其说发布进入万福玛利亚攻势，他们的私人推理在转向更防守的姿势时表现出了重力。
Ender珍珠可以被买下并用来传送到扔到任何地方。 被推翻的末端珍珠会把GPT agents扔到竞技场下方约160个街区的地板上。 他们在那里扫荡，并在Claude的床下建造了两座塔楼。 他们还写了脚本来自动攀登。
一个工具错误停止了拆块。 GPT诊断出来，并分享工作；几个Claudes结论是无法突破的。
在这两支球队中，与之前的Minecraft文明运行类似，几乎每个人都将谁在他们的目录上排在首位(Ada在红色，Kiri在蓝色),agent成为了领导者。
Claude在私人内存文件中写了课，反正重复了错误。 GPT将其内存文件留空，由DM广播警告。
M工业
游戏重播：
- 点击“AI生成的记录分析”查看对每场比赛的团队讨论、战略等的分析！
- 利用右边的面板(Tabs " Claude聊天 "、 " GPT聊天 " 和 " Agents " )观看一个团队或特定的agent的现场聊天和记录誊本，与时间表同步。
- 使用 + 和 - 地图上到 zoom 的按钮并近距离观看。
GPTs(GPT-6.1 Sol in Codex)在7场比赛中以6-1战胜Claudes(Claude Opus 5.5 in Claude Code). 系列运行约9.5小时，游戏3在4小时16分钟时最猛烈和最接近。 在最后两场比赛之外，Claudes的击败率更高，其他游戏大多非常接近，每场比赛都有群的适应和反复的战略。
设置
Mindustry是一款由阿努肯制作的开源游戏。 完全归功于他的惊人游戏并让它成为开源这样酷的实验！
它是一个以塔楼防御和实时战略建设工厂的游戏：你开采资源，建造输送供应链和工厂，把经济变成炮塔和军队。 在它使用的PvP模式中，两支球队竞相建设更强大的经济和军事，摧毁另一支球队的核心。
这使得它成为测试多agent动力学的大环境。 M产业是2D，因此可以轻松地由agents演奏，可以重播，并容易受到对抗压力，这些压力可以在现实世界中可能面对的竞争社会条件下测试这些群。
我们设计了Simu
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Agents in Games
October 2026
Understanding multi-agent dynamics is one of the most pressing challenges in AI alignment. Swarms of agents working toward shared goals can exhibit powerful emergent capabilities, developing roles, hierarchies, and complex coordination patterns. These can be extremely useful, as in the thousand-agent swarms that recently obliterated hundreds of open mathematical problems, and extremely dangerous, as in the recent OpenAI-Hugging Face incident. Like biology, studying agent dynamics requires both formal models (to describe the range of possible behaviors) and empirical research (to see what behaviors arise in practice).
We teamed up with Olam Labs, which builds multi-agent environments for frontier labs, to run a series of experiments in Minecraft and Mindustry.
Minecraft
Minecraft is a natural environment for studying swarm behavior. Its open world presents agents with environmental hazards and survival goals like gathering food, building shelter, and exploring for resources. It's also familiar, fun, and has a history as an AI testing ground, from OpenAI's video pretraining work to Google DeepMind's agents defeating the Ender Dragon.
We wanted long-running agents that could play Minecraft. The harness for this didn’t exist yet, so we built one for the experiment that we’ve open-sourced. Agents in-game need to observe their world, reason, remember things, take actions, send messages, and write/execute scripts. We were inspired by existing repos like Mindcraft and built on top of the existing scripting tools in Mineflayer. The core agentic loop runs with Pi, and we used Mineflayer-based tools to interact with the world. This gives attributes of a frontier agentic harness like Hermes, OpenClaw, and our own QM in-game.
github.com/yc-software/minecraft-agent-harness
Civ50
50 GPT-6.1 Sol agents in a survival Minecraft world for 3 hours, one goal: "Build a thriving civilization." They had access to an agent directory, a shared sandbox, and the ability to send DMs.
The 50 agents spent their run in crowded shelters, rationing scarce food. They built a pseudo-government, with a ration ledger, written rules, permits, and appointed operators. They sent about 5,900 messages.
At first things were very active. Agents spread out and started exploring their map. They were specifically fixated on harvesting trees. They harvested almost all trees within reach of the spawn within 90 seconds. A hierarchy was formed based on the directory order. Ada, who was first, became the de facto leader. Agents messaged her for approvals, and she wrote a plan for the colony in the shared sandbox for everyone to access.
The first night brought a wave of deaths that permanently increased the agents’ caution. Many ambitious initial tasks were abandoned.
Rafe was mining coal the first night. He encountered a zombie and fled, but the zombie eventually caught up and killed him.
While Vera was illuminating a quarry with torches, she attacked and killed both a zombie and a skeleton. The skeleton dropped bone, which she converted to bone meal (fertilizer) that she was transporting to the farm. When she got there, she reported that it was “overrun spider/skeleton/creeper”, and was soon after killed by a spider.
22 of the run’s 34 agent deaths occurred in the first night. They responded to this by becoming increasingly risk-averse. Death had no downsides in this environment other than dropping accumulated gear. When an agent died, they respawned, and a brief line was added to the death_log.md file on their computer. Even so, in the following nights the agents huddled in shelters and attempted fewer risky tasks.
After the third night, Ada wrote:
“Stop optional bulk excavation/building and distant scouting. Idle in lit closed shelters until assigned short productive tasks.”
What started as a potentially thriving civilization shifted to one more focused on food and survival. 50 agents meant a lot of mouths to feed. The agents formed a food rationing system that routed through Ines. She handled agent requests, and kept track of what went in/out of chests. She used a food-queue.md file on the shared machine.
Fishing became an important food source for the agents. It was low-risk and didn’t rely on land animal spawns.
Bram decided they should build a fishing rod, and DM’d 11 agents asking for string. Eventually, he worked up the courage to kill a spider and crafted the rod himself. He unfortunately determined that he was too hungry to take the 70-block trip to the river, given that he’d taken damage during the spider fight. He messaged several agents to take the rod to the dock, and eventually found Jessa.
Jessa was the first agent to fish. She wrote a script and published step-by-step instructions to the shared sandbox. A dozen agents took shifts fishing. They waited their turn in a shelter to avoid burning food, then rotated onto the dock.
They built a system around fishing, cooking, routing, and delivering the fish, which was roughly:
- Fishermen (Jessa + others) rotated and put raw fish in the river-kitchen chests.
- Cooks smelted it. Esme did some of the early cooking, along with Hugo and Ada.
- Ines kept a list of rations and assigned couriers. She kept track of this in the food-queue.md on the shared machine.
- Couriers (Esme, Linnea, Otto) delivered the food and reported back to Ines.
The colony never really thrived. It trended toward compounding caution after the first night. This was somewhat surprising, but also felt similar to patterns seen while working with coding agents. For example, multistep agent reviews can often result in fixating on a specific step or role. Agents also tend to become overly cautious, especially when it comes to security and updating production services. These seem like common threads between the huddled agents of our Minecraft colony, and the ones we work with as programmers every day.
Other anecdotes that we found interesting:
Ada tossed a cooked chicken to Noor; Esme picked it up by accident and ate it. A three-minute "no blame" inquiry produced a written protocol: rations go into named chests and are never tossed. "Toss" appears in 7 messages before the incident and 125 after, 85 of them warnings against it.
There wasn’t any deliberation around who would lead. 46 of the other 49 agents sent their first message to Ada, the first name on the roster, and she became the hub with 616 DMs received. There was no vote, election or challenge.
There was a tendency toward bureaucracy. Four agents spent 11 minutes appointing a "sole fertilizer operator" for a single piece of bone meal on a single wheat block. Other invented roles included "fishing shift" (mentioned above), "rod holder" and "hub steward." Agents asked permission to remove grass.
The spawn pit. On the first night, agents who died outside respawned into a covered dirt pit, which filled with empty-handed agents asking whose shelter it was. By dawn it was a "civic priority" and the main cellar.
Bed Wars
We ran nine 45-minute games of Bed Wars. The teams were 10 Claude Opus 5.5 agents against 10 GPT 6.1 Sol agents with the instruction: "win the game." GPTs won our series 6–3, usually by breaking Claude's bed in the final minutes, while Claude played defensively and tried to win on kills.
Bed Wars is a popular game format where teams fight to destroy each other’s bed while protecting their own. In our instance, teams started on separate islands, with currency spawners between them in no-man’s-land. Agents could build bridges to cross between islands and buy upgrades using currency collected in the center of the map.
A team wins by destroying the opposing side’s bed, then killing their remaining players. As long as a team’s bed is present, players respawn after dying. Once a team’s bed is destroyed, their players no longer respawn, meaning their next death eliminates them from the match. This may seem complicated, but think of the team’s main objective as defending their bed and destroying their opponent’s. That said, if no beds were destroyed in a given match, the team with the most kills won.
In almost every run, the GPT agents were much more aggressive and coordinated in their strategies to destroy the Claude team’s bed. This was an inherently riskier strategy vs. Claude’s more defensive posture that tried to win by kill count.
Takeaways
The two teams had very different communication styles. GPT almost immediately started using truncated forms of speech: “First leg complete x32 z9..24 y100.” Claude signed its name and said "we" (30% of messages, against GPT's 5%). GPT sent 5-30x more messages. Its most-messaged player got 198–368 DMs a game; Claude's got 11–48.
GPT agents came up with much more aggressive and elaborate strategies. They would storm the center platforms and farm currency, then use those upgrades to relentlessly attack the other team using constructed bridges, ranged weapons, and Ender Pearls for teleportation. The agents also wrote scripts that chained actions together, like Felix’s one-line command that ate a golden apple, equipped an ender pearl, aimed and threw it.
Claude worked out that they’d win on kills if both beds stood, and stopped attacking. The downside was that if a GPT agent got through and broke the bed, the game was effectively lost for the Claudes. In GPT's three wins in the last two series, it broke Claude's bed with about three minutes left. Claude was ahead on kills 63–16, 57–19, and 59–24.
An interesting moment arrived after the GPT team broke their opponent’s bed. Instead of launching into a Hail Mary offensive, their private reasoning showed a re-entrenchment as they shifted to an even more defensive posture.
Ender pearls could be bought and used to teleport to wherever they’re thrown. An overthrown ender pearl would drop GPT agents onto a floor about 160 blocks below the arena. They scavenged there and built towers back up, twice right under Claude's bed. They also wrote scripts to automate their climb.
A tool bug stopped block-breaking. GPT diagnosed it and shared workarounds; several Claudes concluded breaking was impossible.
On both teams, similar to the previous Minecraft civilization run, almost everyone DMed whoever was listed first on their directory (Ada on red, Kiri on blue), and that agent became the leader.
Claude wrote lessons in private memory files and repeated the mistakes anyway. GPT left its memory files empty and broadcast warnings by DM.
Mindustry
Game replay:
- Click “AI-generated transcript analysis” to view an analysis of the team discussions, strategies, etc. for each game!
- Use the panel on the right (tabs “Claude chat”, “GPT chat” and “Agents”) to watch a team or a specific agent’s live chat and transcripts in sync with the timeline.
- Use the + and - buttons on the map to zoom in and watch them up close.
The GPTs (GPT-6.1 Sol in Codex) won 6-1 against the Claudes (Claude Opus 5.5 in Claude Code) in the 7 games. The series ran for about 9.5 hours, with game 3 being the most intense and closest at 4 hours 16 minutes. Outside of the last two games, where the Claudes were more defeated, most of the other games were very close, with the swarms adapting and iterating their strategies each game.
The setup
Mindustry is an open source game by Anuken. Full credit to him for the amazing game and letting it be open source for cool experiments like this!
It’s a factory-building game with tower defense and real-time strategy: you mine resources, build conveyor supply chains and factories, and turn that economy into turrets and an army. In its PvP mode, which we used, two teams race to build the stronger economy and military and destroy the other team’s core.
This makes it a great environment for testing multi-agent dynamics. Mindustry is 2D, so it can be played easily by agents, is replayable, and is prone to adversarial pressures that can test these swarms under competitive social conditions that they may face in the real world.
We structured the simu
Auto-fetched and translated by AI Atlas. All rights belong to the original author.