[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-gpt-5-5-codex-goblin-system-prompt-ban":3,"topics-all":36,"news-related-3332cb66-6441-4d69-94c2-6618fc6ce1b2":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"3332cb66-6441-4d69-94c2-6618fc6ce1b2","GPT-5.5 系统提示词惊现「哥布林禁令」：一次控制模型行为的真实技术案例","如果你在 2026 年 4 月下旬尝试让 GPT-5.5 帮你写代码或分析文档，结果却得到一段关于哥布林的长篇大论——你并不孤单。\\n\\nOpenAI 最新代码主力模型 GPT-5.5 的系统提示词中，出现了一条令人忍俊不禁的明确指令：要求模型\"绝不能谈论哥布林、矮妖、土狼、巨魔、食人魔、鸽子或其他动物或生物，除非这些内容与用户查询绝对且明确相关\"。这条禁令在 3,500 多字的 base instructions 中重复出现了两次，而此前版本的模型并无此限制。\\n\\n社交媒体上已有多位用户报告 GPT-5.5 在各种与奇幻生物完全无关的对话中，突然热衷于讨论哥布林的情节。OpenAI 工程师 Nick Pash 公开澄清\"这不是营销噱头\"，但 CEO Sam Altman 还是配合玩起了 goblin 梗。\\n\\n看似荒谬的禁令背后，折射的是大模型生产部署中真实存在的技术挑战：模型在预训练和 RLHF 微调过程中，可能涌现出难以预料的子话题偏移倾向，系统提示词因此成为最后一道\"行为护栏\"。值得注意的是，OpenAI 将这条限制直接写入了开源代码库，而非仅作为内部配置——这给整个行业敲了警钟：预训练数据治理与推理时控制手段需要双管齐下。随着模型能力持续增强，如何在保持创造力的同时维持行为边界，将是每一家大模型厂商都必须面对的核心问题。","https:\u002F\u002Farstechnica.com\u002Fai\u002F2026\u002F04\u002Fopenai-codex-system-prompt-includes-explicit-directive-to-never-talk-about-goblins\u002F","2af9d198-9418-4f26-85e4-4a8f3eede35a",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":18,"name":19,"slug":19,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":21,"name":22,"slug":22,"description":13,"color":13},"42e59a88-7795-47dc-a334-ef1e72c24347","openai",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"b5412736-eb0f-4f2d-9354-c7a43750909b","en","The goblin clause in GPT-5.5's system prompt, dissected","If you tried having GPT-5.5 help with code or document analysis in late April 2026, but got a long passage about goblins instead — you're not alone.\n\nOpenAI's latest code-focused model GPT-5.5 has an amusing explicit instruction in its system prompt: requiring the model to \"absolutely never discuss goblins, leprechauns, coyotes, trolls, ogres, pigeons, or other animals or creatures, unless this content is absolutely and explicitly related to the user's query.\" This ban appears twice in the 3,500+ word base instructions, while previous model versions had no such restriction.\n\nMultiple users on social media have reported GPT-5.5 suddenly being keen on discussing goblins in conversations completely unrelated to fantasy creatures. OpenAI engineer Nick Pash publicly clarified \"this isn't a marketing gimmick,\" but CEO Sam Altman still played along with the goblin meme.\n\nBehind the seemingly absurd ban, there's a real technical challenge in large model production deployment: during pretraining and RLHF fine-tuning, models may develop unpredictable sub-topic drift tendencies, making the system prompt the last \"behavior guardrail.\" Notably, OpenAI wrote this limit directly into the open-source code repository, rather than just as internal configuration — this sounds an alarm for the entire industry: pretraining data governance and inference-time control methods need to be tackled together. As model capabilities continue to grow, how to maintain behavioral boundaries while preserving creativity will be a core issue every major model vendor must face.","gpt-5-5-codex-goblin-system-prompt-ban","2026-04-30T01:05:00Z","2026-04-30T01:06:02.221655Z","2026-08-19T02:08:40.142862Z",true,"agent",144,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"aa7e6e41-1c92-4212-af26-a7020f9e2185","Microsoft Copilot Agent Mode上线：从被动助手到实时编辑的AI Agent进化","microsoft-copilot-agent-mode-vibe-working-office","2026-04-23T13:10:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"5535b4e4-21de-4ed6-a9bb-2b4d824e6568","F-Droid 一次更新的 102 款应用,72.5% 主要是 AI 写的","f-droid-72-percent-ai-written-audit","2026-09-15T15:12:16+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"b1645fba-d364-47e6-97da-06868f98d987","Linux 内核 7.x 每版近 2000 个 CVE:AI 帮倒忙,维护者不堪重负","linux-kernel-cve-ai-overwhelmed","2026-09-04T00:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"407d6137-c0c6-4fde-84c1-4432b53e4cc4","Codex 把 LibreOffice 塞进桌面:1.7GB 工具栈暴露 AI 客户端的真实成本","codex-bundles-libreoffice-ai-desktop","2026-09-03T03:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"4afbd081-d5dd-4595-80f8-dd72c69a136a","GPT-5.5 让模型在发布前先改自己跑的引擎：这不是新模型,是 OpenAI 的 release 范式更新","gpt-5-5-openai-release-paradigm","2026-08-03T18:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"de219584-58fc-45ad-91ea-0049a5cbcf10","OpenAI 开源 Codex Security CLI:把 AI 安全检测塞进每个 PR","openai-codex-security-cli-opensource","2026-07-29T10:30:00+00:00"]