Shared Imagination: LLMs Hallucinate Alike
Abstract: Despite the recent proliferation of large language models (LLMs), their training recipes -- model architecture, pre-training data and optimization algorithm -- are often very similar. This naturally raises the question of the similarity among the resulting models. In this paper, we propose a novel setting, imaginary question answering (IQA), to better understand model similarity. In IQA, we ask one model to generate purely imaginary questions (e.g., on completely made-up concepts in physics) and prompt another model to answer. Surprisingly, despite the total fictionality of these questions, all models can answer each other's questions with remarkable consistency, suggesting a "shared imagination space" in which these models operate during such hallucinations. We conduct a series of investigations into this phenomenon and discuss the implications of such model homogeneity on hallucination detection and computational creativity. We will release and maintain code and data on a public website.
Submission Length: Regular submission (no more than 12 pages of main content)
Assigned Action Editor: ~Gunhee_Kim1
Submission Number: 4206
Loading