Don't Think of the White Bear: Ironic Negation in Transformer Models Under Cognitive Load

TL;DR

This paper investigates how large language models experience ironic rebound when asked to suppress concepts, revealing that suppression often backfires especially with semantic distractors, and identifies neural mechanisms underlying this phenomenon.

Contribution

It introduces systematic experiments and a dataset to analyze ironic rebound in LLMs, linking cognitive phenomena with mechanistic neural insights.

Findings

Rebound occurs immediately after negation and worsens with semantic distractors.

Repetition of content supports better suppression.

Polarity separation predicts rebound persistence.

Abstract

Negation instructions such as 'do not mention $X$' can paradoxically increase the accessibility of $X$ in human thought, a phenomenon known as ironic rebound. Large language models (LLMs) face the same challenge: suppressing a concept requires internally activating it, which may prime rebound instead of avoidance. We investigated this tension with two experiments. \textbf{(1) Load \& content}: after a negation instruction, we vary distractor text (semantic, syntactic, repetition) and measure rebound strength. \textbf{(2) Polarity separation}: We test whether models distinguish neutral from negative framings of the same concept and whether this separation predicts rebound persistence. Results show that rebound consistently arises immediately after negation and intensifies with longer or semantic distractors, while repetition supports suppression. Stronger polarity separation correlates with more persistent rebound. Together, these findings, complemented by a circuit tracing analysis that identifies sparse middle-layer attention heads amplifying forbidden tokens while early layers suppress, link cognitive predictions of ironic rebound with mechanistic insights into long-context interference. To support future work, we release ReboundBench, a dataset of $5,000$ systematically varied negation prompts designed to probe rebound in LLMs.