不给提示时，大模型会自发生成特定内容，暴露其内在偏好。
What LLMs Think When You Don't Tell Them What to Think About?
- 用无主题输入测试模型，观察其自发生成行为
- 不同模型家族有明显内容偏好，如GPT-OSS偏编程数学
- 发现模型生成退化现象，如Llama输出个人社交链接
研究大语言模型在无明确主题输入下的自发生成行为，揭示其潜在的系统性偏好。尽管输入无主题，各模型仍覆盖广泛语义空间，且不同模型家族表现出显著内容倾向：GPT-OSS主要生成编程（27.1%）和数学内容（24.6%），Llama多生成文学内容（9.1%），DeepSeek常生成宗教内容，Qwen则高频生成选择题。此外，模型在内容深度上存在差异，GPT-OSS更倾向生成动态规划等高级技术内容，而其他模型以基础Python代码为主。研究还发现，无约束生成易导致重复性输出，例如Llama生成多个指向个人Facebook和Instagram账号的链接。本文发布包含16个模型共25.6万条样本的完整数据集及可复现代码库。
原文摘要 · Abstract (English)
Characterizing the behavior of large language models (LLMs) across diverse settings is critical for reliable monitoring and AI safety. However, most existing analyses rely on topic- or task-specific prompts, which can substantially limit what can be observed. In this work, we study what LLMs generate from minimal, topic-neutral inputs and probe their near-unconstrained generative behavior. Despite the absence of explicit topics, model outputs cover a broad semantic space, and surprisingly, each model family exhibits strong and systematic topical preferences. GPT-OSS predominantly generates programming (27.1%) and mathematical content (24.6%), whereas Llama most frequently generates literary content (9.1%). DeepSeek often generates religious content, while Qwen frequently generates multiple-choice questions. Beyond topical preferences, we also observe differences in content specialization and depth: GPT-OSS often generates more technically advanced content (e.g., dynamic programming) compared with other models (e.g., basic Python). Furthermore, we find that the near-unconstrained generation often degenerates into repetitive phrases, revealing interesting behaviors unique to each model family. For instance, degenerate outputs from Llama include multiple URLs pointing to personal Facebook and Instagram accounts. We release the complete dataset of 256,000 samples from 16 LLMs, along with a reproducible codebase.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。