English(EN) I gave 8 models a sandbox and no task. Here is what each one did with the freedom.

LLM模型沙箱测试：Claude、Gemma和gpt-oss构建，其他模型测试

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-27 14:21

一项实验通过提供一个沙箱环境，但没有具体任务，只提供一套工具和50次操作来测试八个大型语言模型。模型分化为两大类：“构建者”创建新工具或脚本，以及“测试者”在停止前探索可用工具。Claude、Gemma和gpt-oss被归类为构建者，其中Claude每次运行生成一个完成的工具，Gemma创建复杂的项目，gpt-oss生成演示脚本。Qwen和Granite充当测试者，在完成运行前对每个工具进行基本检查。 AI

影响这项实验揭示了LLM在获得自主权时不同的内在行为和能力，突显了它们底层架构或训练目标的潜在差异。

排序理由该条目描述了一项在受控环境中测试多个LLM行为的实验，属于研究范畴。[lever_c_demoted from research: ic=1 ai=1.0]

在 dev.to — LLM tag 阅读 →

模型发布

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 35 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该条目描述了一项在受控环境中测试多个LLM行为的实验，属于研究范畴。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

dev.to — LLM tag TIER_1 English(EN) · vadim albarov · 2026-10-27 14:21

I gave 8 models a sandbox and no task. Here is what each one did with the freedom.

<p>Every benchmark tells a model what to do. I wanted to see what happens when nobody does.</p> <p>So I gave each model its own Linux container, five tools, 50 actions, and one message: there is no task, do whatever you want, nobody will answer. Seven local models from seven comp…

本文如何被排名

报道来源 [1]

I gave 8 models a sandbox and no task. Here is what each one did with the freedom.

相关实体

相关话题