What LLMs Think When You Don't Tell Them What to Think About?

TL;DR

This study investigates the intrinsic generative behaviors of various large language models when given minimal, neutral prompts, revealing their topical preferences, content specialization, and unique degeneration patterns.

Contribution

It provides a comprehensive analysis of LLMs' unconstrained output behaviors, highlighting systematic topical biases and content characteristics across multiple models.

Findings

Models exhibit strong and systematic topical preferences.

Content depth and specialization vary significantly between models.

Degeneration patterns include repetitive phrases and model-specific behaviors.

Abstract

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.