The Expense of Seeing: Attaining Trustworthy Multimodal Reasoning Within the Monolithic Paradigm

TL;DR

This paper critically examines the trustworthiness of current Vision-Language Models, proposing a new information-theoretic framework to measure and improve genuine multimodal reasoning beyond dataset biases.

Contribution

It introduces the Modality Translation Protocol and three novel metrics to quantify the visual knowledge bottleneck, challenging existing evaluation methods and guiding future model design.

Findings

State-of-the-art models often exploit language priors, bypassing visual data.

The proposed metrics reveal the expense and limitations of current visual reasoning.

Scaling language models may increase visual knowledge bottleneck penalties.

Abstract

The rapid proliferation of Vision-Language Models (VLMs) is often framed as enabling unified multimodal knowledge discovery but rests on an under-examined assumption: that current VLMs faithfully synthesise multimodal data. We argue they often do not, and this gap reflects a trustworthiness problem in the dominant Vision Encoder-Projector-LLM paradigm. Rather than extracting grounded knowledge from visual inputs, state-of-the-art models frequently exhibit functional blindness, i.e., exploiting strong language priors to bypass severe visual representation bottlenecks. In this work, we challenge the conventional methodology of multimodal evaluation, which relies on data ablation or new dataset creation and therefore conflates dataset biases with architectural incapacity. We propose an information-theoretic departure: the Modality Translation Protocol, designed to quantify what we call the Expense of Seeing. By translating semantic payloads rather than ablating them, we formulate three novel metrics -- the Toll (ToS), Curse (CoS), and Fallacy (FoS) of Seeing -- culminating in the Semantic Sufficiency Criterion (SSC). Furthermore, we hypothesise a Divergence Law of Multimodal Scaling: as the underlying language engines scale to unprecedented reasoning capabilities, the penalty of the visual knowledge bottleneck may increase rather than diminish. We argue the community should move beyond "multimodal gain" as a primary evaluation target. By elevating the SSC from a passive diagnostic constraint to an active architectural blueprint, we provide a foundation for guiding the next generation of AI systems toward genuine multimodal reasoning.