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

Hugging Face Daily Papers Papers
omni-modal-llms modality-bias evidence-conflict benchmark cross-modal ai-research

Summary

The paper introduces Tri-PvP, a benchmark that exposes visual bias and asymmetric evidence-form preferences in omni-modal large language models, revealing deep-seated modality biases that are linearly decodable from early layers and resistant to surface mitigation.

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.
Original Article
View Cached Full Text

Cached at: 09/23/26, 03:33 PM

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

Source: https://huggingface.co/papers/2609.06011

Abstract

Tri-PvP benchmark reveals visual bias and asymmetric evidence-form preferences in omni-modal language models, with early-layer decodable modality effects resistant to surface mitigation.

Omni-modal large language models(OLLMs) jointly process vision, audio, and text, yet their modality bias undercross-modal conflictremains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality:perceptual signals(e.g., a photograph or recording of a dog) andpropositional signals(e.g., the declarative claim “this is a dog”), such that any measured modality bias is inherently confounded withevidence-form bias, precluding clean attribution to either source. To address this, we introduceTri-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 robustvisual biasacross most models and evidence-type conditions. Crucially, we reveal a systematic asymmetry inevidence-form bias: models exhibit a stronger bias toward perceptual signal in vision but propositional in audio. Further analyses vialayer-wise linear probingandcontrastive decodingreveal 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.

View arXiv pageView PDFGitHub2Add to collection

Get this paper in your agent:

hf papers read 2609\.06011

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.06011 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.06011 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.06011 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

OPD-V: Visual On-Policy Self-Distillation with Modality Balance

Hugging Face Daily Papers

Introduces OPD-V, a visual on-policy self-distillation paradigm for multimodal large language models that leverages positive and negative teachers to exploit modality balance as privileged information, improving reasoning performance across benchmarks while reducing training cost.