mlsanityData to 7 Oct 2026

A Shared Taste for Model-Written Text: The Generator-by-Selector Matrices of "AI-AI Bias" Show No Detectable Own-Model Premium

arXiv 2610.00369 ↗First posted 30 Sep 2026cs.IR cs.AI cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Laurito et al. (PNAS 2025) showed that large language models choosing between two descriptions of the same product, paper or film prefer the description written by a language model over the one written by a person, by a wide margin over what human judges do. Their design crosses five generators with the same five models as selectors, which permits a second question the paper does not headline: does a selector prefer text from its own model beyond what the generator and selector main effects predict? We rebuild the three 5x5 matrices from the per-item counts in the authors' public repository (21,828 valid trials; every cell matches the published value) and fit a two-way fixed-effects model with an own-model term gamma, tested by the exact permutation test over the 120 relabellings of the selectors. The premium is +0.013 on products (exact one-sided p = 0.24), -0.010 on paper abstracts (p = 0.74), +0.054 on films (p = 0.07) and +0.019 pooled (p = 0.14; 95% interval -0.008 to 0.046). The same-vendor term for the GPT-3.5 and GPT-4 pair is negative in all three datasets. Position bias moves single cells by up to 0.42 share points in either direction, and the own-model contrast is unchanged once order-driven items are removed. The design would have detected a premium of 0.05 with 82% (products), 88% (papers), 42% (films) and 97% (pooled) power; the minimum detectable effect at 80% power is 0.034 pooled. The absence is informative down to about 0.04 share points and silent below that. The 4x4 matrix of Tan et al. (ACL 2024) gives gamma = +0.148 at the smallest p its 24 relabellings allow, with a same-family term of the same size. The main result of Laurito et al. stands: models share a taste for model-written text, with GPT-4's descriptions chosen 77% to 95% of the time by every selector on products. What these data do not show is a model recognising and favouring its own prose.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 21 of its 25 references that are in the map. First posted on arXiv 30 Sep 2026.

Builds on

21 of 25 references.

References · 21

  1. 2024LLM Evaluators Recognize and Favor Their Own Generationscited by 651
  2. 2024Self-Preference Bias in LLM-as-a-Judgecited by 248
  3. 2025Preference Leakage: A Contamination Problem in LLM-as-a-judgecited by 119
  4. 2024Blinded by Generated Contexts: How Language Models Merge Generated and Retrieved Contexts When Knowledge Conflicts?cited by 47
  5. 2025Do LLM Evaluators Prefer Themselves for a Reason?cited by 38
  6. 2025Play Favorites: A Statistical Method to Measure Self-Bias in LLM-as-a-Judgecited by 35
  7. 2025Beyond the Surface: Measuring Self-Preference in LLM Judgmentscited by 30
  8. 2024AI-AI Bias: large language models favor communications generated by large language modelscited by 25
  9. 2024Self-Recognition in Language Modelscited by 17
  10. 2025What Generative Search Engines Like and How to Optimize Web Content Cooperativelycited by 13
  11. 2025AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insightscited by 10
  12. 2026Self-Preference Bias in Rubric-Based Evaluation of Large Language Modelscited by 9
9 more
  1. 2026Quantifying and Mitigating Self-Preference Bias of LLM Judgescited by 9
  2. 2025Know Thyself? On the Incapability and Implications of AI Self-Recognitioncited by 7
  3. 2024LLMs are Biased Evaluators But Not Biased for Retrieval Augmented Generationcited by 5
  4. 2026Are LLM Evaluators Really Narcissists? Sanity Checking Self-Preference Evaluationscited by 5
  5. 2026Self-Preference Is Weak or Absent in Verifiable Instruction-Following Revision: A Four-Model Test Under Genuine Authorshipcited by 2
  6. 2026Evaluating Large Language Model Raters for German Open-Response Clinical Questions: A Physician-Annotated Benchmark Study of Agreement, Evaluator Bias, and Abstentioncited by 2
  7. 2026On the impact of retrieved content representations in RAG Pipelinescited by 1
  8. 2026Self- and Other-Labels Induce Bidirectional Bias in LLM Judgescited by 1
  9. 2026Who Judges Matters: Measuring Family-Conditioned Preference in LLM-as-Judge Panelscited by 1

Built on by

Not cited yet.