Newest papers

What Language Adds, and at What Cost: Four Controlled Probes of Pathology Vision-Language Models

Wenhao Zhang, Zhongliang Zhou, Yi Ping, Zirui Wang, Sheng Li · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26805 · cited by 0 works (OpenAlex) · open-access PDF

German

Vision-Agent-Sprachmodelle (VLMs) versprechen, die computergestützte Pathologie nachvollziehbar und in der Lage zu machen, Berichte zu erstellen, aber wie viel Sprache über den Rückenmark der Vision hinzufügt, ist unklar, weil fast jedes System von seinen Entwicklern auf Testsätzen ihrer eigenen Wahl bewertet wird. Wir bewerten die veröffentlichten Beweise nach der Art und Weise, wie sie produziert wurden und testen vier Behauptungen mit kontrollierten Sonden, die veröffentlichte Checkpoints und öffentliche Daten wiederverwenden, wobei eine Eingabe zu einer Zeit variiert. Berichtstext fügt ein prognostisches Signal hinzu: das Hinzufügen einer eingefrorenen Einbettung des am Ende bearbeiteten Berichts zu einem Slide-Einbett

Machine translation (Meta M2M100); the English original follows.

Abstract

Vision--language models (VLMs) promise to make computational pathology queryable and able to draft reports, but how much language adds beyond the vision backbone is unclear, because almost every system is evaluated by its developers on test sets of their own choosing. We grade published evidence by how it was produced and test four claims with controlled probes that reuse released checkpoints and public data, varying one input at a time. Report text adds prognostic signal: adding a frozen embedding of the stage-redacted report to a slide embedding raises the pooled C-index by 0.019 for progression-free interval and by 0.016 for overall survival across 1{,}633 TCGA patients, and restoring the stage statement adds nothing. Prompts add no measurable accuracy at inference: with features and pooling fixed, zero-shot classification is statistically indistinguishable from a supervised linear probe for lung and renal cancer subtyping. Slide-level question answering is barely grounded: SlideChat answers 74.82\% of SlideBench-VQA(TCGA) items correctly with the slide and 74.41\% without it, whereas its language backbone alone reaches 43.52\%. Paraphrasing leaves accuracy unchanged but flips 8--23\% of answers. Self-reported gains from agentic navigation and few-shot language priors cannot yet be attributed to language. Language thus adds real but narrow value, and we propose minimum controls that would expose these failure modes in future evaluations.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Pooling, Computer science, Natural language processing, Reuse, Embedding, Artificial intelligence

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange