mlsanityData to 13 Oct 2026

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

arXiv 2511.12381 ↗First posted 15 Nov 2025cs.CL cs.AIcited by 1

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

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. (1) Load & content: after a negation instruction, we vary distractor text (semantic, syntactic, repetition) and measure rebound strength. (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.

Key facts. Cited by 1 paper in the map (data through 13 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 12 of its 24 references that are in the map. First posted on arXiv 15 Nov 2025.

Builds on

12 of 24 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025Qwen3 Technical Report14 May
  2. 2025Negation: A Pink Elephant in the Large Language Models' Room?28 Mar

Other references · 10

  1. 2024The Llama 3 Herd of Modelscited by 14,486
  2. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,500
  3. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,412
  4. 2022BLOOM: A 176B-Parameter Open-Access Multilingual Language Modelcited by 1,978
  5. 2025gpt-oss-120b & gpt-oss-20b Model Cardcited by 1,319
  6. 2023Copy Suppression: Comprehensively Understanding an Attention Headcited by 75
  7. 2024Investigating and Addressing Hallucinations of LLMs in Tasks Involving Negationcited by 15
  8. 2024Mitigating Copy Bias in In-Context Learning through Neuron Pruningcited by 11
  9. 2025Thunder-NUBench: A Benchmark for LLMs' Sentence-Level Negation Understandingcited by 5
  10. 2024Do not think about pink elephant!cited by 3

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Can You Keep a Secret? Involuntary Information Leakage in Language Model Writingcited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Interpreting Negation in GPT-2: Layer- and Head-Level Causal Analysis
  2. 2025Disentangling Recall and Reasoning in Transformer Models through Layer-wise Attention and Activation Analysis
  3. 2026Lexical Perturbations Disrupt LLM Reasoning: An Empirical Study of Attention Diversion
  4. 2026Thinking Inertia: LLMs Keep Thinking When Told Not To
  5. 2026The First Drop of Ink: Nonlinear Impact of Distracting Information in Long-Context Reasoning
  6. 2025Erased or Dormant? Rethinking Concept Erasure Through Reversibility
  7. 2026The Spectral Geometry of Thought: Phase Transitions, Instruction Reversal, Token-Level Dynamics, and Perfect Correctness Prediction in How Transformers Reason
  8. 2023The Hydra Effect: Emergent Self-repair in Language Model Computations