Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts

Yenting, Chen Shu Yun, Chin Hui Chu, Chun-Wei Chen, Shih-Yun Shan Kuan, Hung-yi Lee, Yun-Nung Chen

Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts: 18 upvotes on Hugging Face Daily Papers, #16 of 24 papers on 2026-09-23. Day-by-day upvote history.

Omni-modal large language models (OLLMs) jointly process vision, audio, and text, yet their modality bias under cross-modal conflict remains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality: perceptual signals (e.g., a photograph or recording of a dog) and propositional signals (e.g., the declarative claim "this is a dog"), such that any measured modality bias is inherently confounded with evidence-form bias, precluding clean attribution to either source. To address this, we introduce Tri-PvP, an 8,000-sample tri-modal conflict benchmark crossing vision, audio, and text, where vision and audio each take perceptual or propositional form. Evaluating five OLLMs, we find robust visual bias across most models and evidence-type conditions. Crucially, we reveal a systematic asymmetry in evidence-form bias: models exhibit a stronger bias toward perceptual signal in vision but propositional in audio. Further analyses via layer-wise linear probing and contrastive decoding reveal that modality bias is already linearly decodable from early representation layers and can only be partially mitigated, calling for mitigation strategies beyond surface-level interventions.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.