# Where the goblins came from

Canonical URL: https://www.traeai.com/articles/7517fb1b-e89b-46c8-b550-8a22ad4a027d
Original source: https://openai.com/index/where-the-goblins-came-from
Source name: OpenAI Blog
Content type: article
Language: 英文
Score: 8.0
Reading time: 6 分钟
Published: 2026-04-29T20:00:00+00:00
Tags: OpenAI, GPT-5.1, 自然语言处理

## Summary

OpenAI发现GPT-5.1及后续版本中频繁出现“小妖精”等比喻，最终追溯到Nerdy个性设置中的奖励机制。

## Key Takeaways

- GPT-5.1开始频繁使用‘小妖精’等比喻。
- 问题根源在于Nerdy个性设置中的奖励机制。
- 模型行为受多种微小激励影响。

## Outline

- 引言 — 介绍GPT-5.1及后续版本中出现的‘小妖精’比喻现象及其背景。
  - 第一个迹象 — 描述首次发现‘小妖精’比喻的时间和用户反馈情况。
  - 解决谜团 — 详细说明如何通过分析找到问题根源，并解释Nerdy个性设置的影响。

## Highlights

- > Starting with GPT-5.1, our models began developing a strange habit: they increasingly mentioned goblins, gremlins, and other creatures in their metaphors. — 第 2 段
- > A single “little goblin” in an answer could be harmless, even charming. Across model generations, though, the habit became hard to miss: the goblins kept multiplying, and we needed to figure out where — 第 3 段
- > The short answer is that model behavior is shaped by many small incentives. In this case, one of those incentives came from training the model for the personality customization feature, in particular — 第 4 段

## Citation Guidance

When citing this item, prefer the canonical traeai article URL for the AI-readable summary and include the original source URL when discussing the underlying source material.