AI觉醒星球
Awakening is here
Knowledge File / AI技能杠杆
2026-06-07 0 浏览 会员

趋势解读:Five labs,five minds,解读最新 AI 进展

本文介绍了第二届构建小型黑客马拉松的工程报告,探讨了在多代理经济模拟中运行不同实验室的小型模型(异构模型)的实践,包括技术挑战、服务层瓶颈以及游戏设计中的安全考虑。

SOURCE / AI技能杠杆 MIN / 9 ACCESS / 会员 POST / 2026-06-07 03:02:33

原贴

查看原文
作者:Hugging Face Blog 来源站点:huggingface.co 原贴时间:

原文

A second Build Small Hackathon field report: what happens when each agent in an emergent economy runs on a different lab's small model, and the player becomes the financier pulling the strings. The first version of Thousand Token Wood was a weather-god sandbox: five woodland creatures on one fine-tuned 0.5B model traded goods, and you poked the world with shocks and watched bubbles and crashes emerge. It was a nice toy. It was also something you watched rather than played. v2 rebuilt it into a game you operate. You are the Patron of the Wood, a shadow financier: you lend at interest, whisper tips that may be true or planted, short the market, bribe, and broker alliances, while a magistrate hunts you for trading on what you should not know. The creatures remember how you treated them and scheme back. And the biggest change is under the hood: every creature now thinks with a different lab's small model. This is the engineering report. The obvious way to run a council of agents is one model, many prompts. v2 runs four: gpt-oss-20b (OpenAI), MiniCPM3-4B (OpenBMB), Nemotron-Mini-4B (NVIDIA), and a fine-tuned Qwen 0.5B of my own. The point is not novelty for its own sake. A market is interesting when the participants genuinely differ, and four labs' models trained on different data with different post-training are about as different as small models get. The owl hoards differently than the fox speculates. The council is a live argument, not a script. Standing four distinct models up on one platform surfaced the real lesson: the friction is almost entirely at the serving layer, not the modeling layer. Current vLLM (0.22.1) JIT-compiles kernels at load and needs the CUDA toolkit ( nvcc ) present. A lean base image does not ship it, so all four models failed identically with "could not find nvcc" until I based them on a CUDA devel image. This was not a gpt-oss quirk; it was universal to the vLLM version. One image fix unblocked everything. gpt-oss-20b runs in its native MXFP4 quantization and fits a 24GB L4 with room to spare; no high-end GPU needed. It also speaks a channel format that wraps the answer in an analysis preamble, so the consumer has to extract the final channel. MiniCPM3 needed trust_remote_code ; Nemotron loaded clean. Per-model footguns, each a one-line config. The thing that made four heterogeneous models tractable was the same primitive that made one model tractable in v1: a tolerant JSON parse-and-repair layer that every model's output flows through. Different tokenizers and formatting habits produce different malformations; the parser drops what it cannot salvage and the simulation never crashes. Build that layer once and adding a model is a config entry, not a refactor. The dramatic core of v2 is the insider tip. You can whisper a tip to a creature that is true (a real forecast of the next market mania the deck will draw, your genuine edge) or false (bait). Acting on a true tip and profiting raises your heat; cross a threshold and the magistrate opens an investigation that ends in a fine, frozen assets, or exile. For that to be a real game, the truth of a tip must be hidden from the creatures. They see the rumor text; they must never see the flag. This is a security property, not a UI nicety, and small-model agents make it sharp: everything the model could repeat back is whatever you put in its prompt. So the hidden flag lives off-prompt entirely (on the player's ledger), it is stripped from the public event record at construction, and the only thing the narrator ever summarizes is public events. A single test scans every creature's full prompt, every turn, for the banned tokens. That test is the most important one in the suite. When you give an agent secret information, assume it will leak unless a test proves it cannot. Creatures carry persistent relationships: a signed sentiment toward the Patron and toward each other, nudged by events (you shorted my crop, you repaid your loan, you allied me with a r

中文翻译

第二届构建小型黑客马拉松现场报告:当新兴经济中的每个代理运行在不同实验室的小模型上,而玩家成为幕后操纵的金融家时会发生什么。千字森林的第一个版本是一个天气之神沙盒:五个林地生物在一个微调的0.5B模型上交易商品,你通过冲击戳击世界,观察泡沫和崩溃的出现。它是一个不错的小玩具。但它也是你观看而不是玩耍的东西。v2被重建为你操作的游戏。你是森林的守护者,一个影子金融家:你以利息放贷,耳语可能是真实或植入的提示,做空市场,贿赂,撮合联盟,同时一名治安官追捕你,因为你交易了你不应该知道的东西。生物们记得你如何对待它们,并阴谋反击。最大的变化在引擎盖下:每个生物现在用不同实验室的小模型思考。这是一份工程报告。运行代理理事会的显而易见的方式是一个模型,多个提示。v2运行四个:gpt-oss-20b (OpenAI), MiniCPM3-4B (OpenBMB), Nemotron-Mini-4B (NVIDIA), 以及我自己的微调Qwen 0.5B。目的不是为了新奇而新奇。当参与者真正不同时,市场才有趣,四个实验室的模型在不同数据上训练,采用不同的后训练,它们之间就像小模型能有的那么不同。猫头鹰囤积的方式与狐狸投机的不同。理事会是一个实时争论,而不是一个脚本。将四个不同模型架设在一个平台上揭示真正的教训:摩擦几乎完全在服务层,而不是模型层。当前的vLLM (0.22.1) 在加载时即时编译内核,需要CUDA工具包 (nvcc) 存在。一个精简的基础镜像没有携带它,所以所有四个模型都因“找不到nvcc”而失败,直到我基于CUDA开发镜像。这不是gpt-oss的怪癖;这是vLLM版本的普遍问题。一个镜像修复解除了所有障碍。gpt-oss-20b以其原生MXFP4量化运行,并可装入24GB L4且有余量;不需要高端GPU。它还说一种通道格式,将答案包裹在分析前导中,因此消费者必须提取最终通道。MiniCPM3需要trust_remote_code;Nemotron干净加载。每个模型的陷阱,每个都是一行配置。使四个异构模型可管理的与使一个模型在v1中可管理的原语相同:一个宽容的JSON解析和修复层,每个模型的输出都流经它。不同的分词器和格式化习惯产生不同的畸形;解析器丢弃无法修复的内容,模拟从不崩溃。构建该层一次,添加一个模型就是一个配置条目,而不是重构。v2的核心戏码是内幕提示。你可以向生物耳语一个真实提示(对下一轮市场狂热的真实预测,你的真正优势)或虚假提示(诱饵)。根据真实提示行动并获利会提高你的热度;越过阈值,治安官展开调查,最终导致罚款、资产冻结或流放。为了使其成为真正的游戏,提示的真假必须对生物隐藏。它们看到谣言文本;它们绝不能看到标志。这是一个安全属性,而不是UI的精细之处,小模型代理使这一点变得尖锐:模型能复述的一切就是你放入其提示中的内容。所以隐藏标志完全远离提示(在玩家的账本上),它在构造时从公共事件记录中剥离,叙述者总结的只有公共事件。一个测试在每一轮扫描每个生物的完整提示中是否存在被禁止的令牌。这个测试是套件中最重要的一个。当你给代理秘密信息时,假设它会泄露,除非测试证明不能。生物持有持久的关系:对守护者和彼此之间的带符号情感,由事件推动(你做空了我的庄稼,你还了贷款,你与我结盟

核心信息

本文介绍了第二届构建小型黑客马拉松的工程报告,探讨了在多代理经济模拟中运行不同实验室的小型模型(异构模型)的实践,包括技术挑战、服务层瓶颈以及游戏设计中的安全考虑。

  • 多模型异构代理模拟更真实的市场行为
  • 服务层(vLLM)是主要技术摩擦点
  • 建立容错JSON解析层可简化多模型集成
  • 隐藏秘密信息需彻底从提示中剥离并测试
  • 每个模型需单独配置但整体可管理
试看内容

成为会员查看完整内容

你已经看到了这篇内容的前置整理,剩余深度部分仅对会员开放。

详细解读 信息差价值 参考来源
成为会员查看完整内容
上一篇 趋势解读:ChatGPT's new Lockdown Mode lets you disable web,提升开发者接入体验 下一篇 趋势解读:Sakana AI bets AI that improves itself can,解读最新研究结论