As LLM-as-a-judge becomes increasingly widespread, self-preference—the tendency of a judge to favor its own outputs—raises growing concerns about evaluation reliability. However, this bias has been studied predominantly on generated text, where stylistic features and response quality are inevitably conflated. As a result, existing measurements cannot separate genuine self-preference from these confounds. We address this limitation by changing the object of evaluation: instead of judging generated text, ten LLMs assess sets of narrative constraints selected from a shared pool, which carry no stylistic fingerprint yet retain a recoverable model-specific signature. Two experiments on this task yield complementary findings. Under blind evaluation, self-preference disappears, with a small effect remaining in the opposite direction once selection quality and judge severity are controlled. Under matched quality, however, self- and other-labels alone—without naming any model—shift scores bidirectionally. LLM judges inflate scores for self-labeled selections and deflate those for other-labeled ones regardless of the selection’s actual source. We make two contributions: 1) authorship attribution is a distinct driver of evaluation bias, and 2) ground-truth-free tasks can serve as controlled instruments for studying LLM judge behavior.

This paper describes the methods used for the NAACL 2024 workshop MEDIQA-M3G shared task for generating medical answers from image and query data for skin diseases. MedVInT-Decoder, LLaVA, and LLaVA-Med are chosen as base models. Finetuned with the task dataset on the dermatological domain, MedVInT-Decoder achieved a BLEU score of 3.82 during competition, while LLaVA and LLaVA-Med reached 6.98 and 4.62 afterward, respectively.