On the Cone Effect and Modality Gap in Medical Vision-Language Embeddings

• URL: http://arxiv.org/abs/2603.17246v2
• Date: Fri, 20 Mar 2026 15:22:06 GMT
• Title: On the Cone Effect and Modality Gap in Medical Vision-Language Embeddings
• Abstract summary: Vision-Language Models (VLMs) exhibit a characteristic "cone effect" in which nonlinear encoders map embeddings into highly concentrated regions of the representation space.<n>We introduce a lightweight post-hoc mechanism that keeps pretrained VLM encoders frozen while continuously controlling cross-modal separation.<n>This enables systematic analysis of how the modality gap affects downstream multimodal performance without expensive retraining.
• Score: 6.66266390476704
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Vision-Language Models (VLMs) exhibit a characteristic "cone effect" in which nonlinear encoders map embeddings into highly concentrated regions of the representation space, contributing to cross-modal separation known as the modality gap. While this phenomenon has been widely observed, its practical impact on supervised multimodal learning -- particularly in medical domains -- remains unclear. In this work, we introduce a lightweight post-hoc mechanism that keeps pretrained VLM encoders frozen while continuously controlling cross-modal separation through a single hyperparameter {λ}. This enables systematic analysis of how the modality gap affects downstream multimodal performance without expensive retraining. We evaluate generalist (CLIP, SigLIP) and medically specialized (BioMedCLIP, MedSigLIP) models across diverse medical and natural datasets in a supervised multimodal settings. Results consistently show that reducing excessive modality gap improves downstream performance, with medical datasets exhibiting stronger sensitivity to gap modulation; however, fully collapsing the gap is not always optimal, and intermediate, task-dependent separation yields the best results. These findings position the modality gap as a tunable property of multimodal representations rather than a quantity that should be universally minimized.

Related papers

• Diagnosing and Mitigating Perception-Decision Misalignment in Omni-LLMs via Modality Subspace Activation [32.16800432913683]
  We formalize Causal Modality Sensitivity (CMS) operationalized via a dual-lens framework.<n>We also curate CausalMSBench, a diagnostic dataset isolating language priors.<n>We propose Modality Subspace Activation (MSA), a training-free inference-time framework.
  arXiv  Detail & Related papers  (2026-07-31T05:46:40Z)
• Resilient Vision-Tabular Multimodal Learning under Modality Missingness [1.696842238811138]
  Multimodal deep learning has shown strong potential in medical applications.<n>Most existing approaches implicitly assume complete modality availability.<n>We propose a multimodal transformer framework for joint vision-tabular learning.
  arXiv  Detail & Related papers  (2026-05-12T12:14:13Z)
• M-IDoL: Information Decomposition for Modality-Specific and Diverse Representation Learning in Medical Foundation Model [53.96788246923603]
  We propose M-IDoL, a self-supervised underlinetextitMFM that introduces Information Decomposition for multimodal representation Learning.<n>By pre-training on 1.15 million medical images, M-IDoL i) delivers superior generalization across 21 downstream clinical tasks.
  arXiv  Detail & Related papers  (2026-04-10T04:06:11Z)
• Uncertainty-Aware Vision-Language Segmentation for Medical Imaging [12.545486211087791]
  We introduce a novel uncertainty-aware multimodal segmentation framework for medical diagnosis.<n>We propose a Modality Decoding Attention Block (MoDAB) with a lightweight State Space Mixer (SSMix) to enable efficient cross-modal fusion.<n>Our results highlight the importance of incorporating uncertainty modelling and structured modality alignment in vision-language medical segmentation tasks.
  arXiv  Detail & Related papers  (2026-02-16T06:27:51Z)
• Explaining multimodal LLMs via intra-modal token interactions [55.27436637894534]
  Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse vision-language tasks, yet their internal decision-making mechanisms remain insufficiently understood.<n>We propose enhancing interpretability by leveraging intra-modal interaction.
  arXiv  Detail & Related papers  (2025-09-26T14:39:13Z)
• Multimodal Causal-Driven Representation Learning for Generalizable Medical Image Segmentation [56.52520416420957]
  We propose Multimodal Causal-Driven Representation Learning (MCDRL) to tackle domain generalization in medical image segmentation.<n>MCDRL consistently outperforms competing methods, yielding superior segmentation accuracy and exhibiting robust generalizability.
  arXiv  Detail & Related papers  (2025-08-07T03:41:41Z)
• Towards Scalable and Robust White Matter Lesion Localization via Multimodal Deep Learning [2.0749231618270803]
  White matter hyperintensities (WMH) are radiological markers of small vessel disease and neurodegeneration, whose accurate segmentation and localization are crucial for diagnosis and monitoring.<n>We propose a deep learning framework for WM lesion segmentation and localization that operates directly in native space using single- and multi-modal MRI inputs.<n>Our findings highlight the utility of multimodal fusion for accurate and robust WMH analysis, and the potential of joint modeling for integrated predictions.
  arXiv  Detail & Related papers  (2025-06-27T09:39:26Z)
• Reducing Unimodal Bias in Multi-Modal Semantic Segmentation with Multi-Scale Functional Entropy Regularization [66.10528870853324]
  Fusing and balancing multi-modal inputs from novel sensors for dense prediction tasks is critically important.<n>One major limitation is the tendency of multi-modal frameworks to over-rely on easily learnable modalities.<n>We propose a plug-and-play regularization term based on functional entropy, which introduces no additional parameters.
  arXiv  Detail & Related papers  (2025-05-10T12:58:15Z)
• Robust Multimodal Learning for Ophthalmic Disease Grading via Disentangled Representation [30.697291934309206]
  multimodal data is rare in real-world applications due to a lack of medical equipment and concerns about data privacy.<n>Traditional deep learning methods typically address these issues by learning representations in latent space.<n>Authors propose the Essence-Point and Disentangle Representation Learning (EDRL) strategy, which integrates a self-distillation mechanism into an end-to-end framework.
  arXiv  Detail & Related papers  (2025-03-07T10:58:38Z)
• Mitigating Hallucination for Large Vision Language Model by Inter-Modality Correlation Calibration Decoding [66.06337890279839]
  Large vision-language models (LVLMs) have shown remarkable capabilities in visual-language understanding for downstream multi-modal tasks.<n>LVLMs still suffer from generating hallucinations in complex generation tasks, leading to inconsistencies between visual inputs and generated content.<n>We propose an Inter-Modality Correlation Decoding (IMCCD) method to mitigate hallucinations in LVLMs in a training-free manner.
  arXiv  Detail & Related papers  (2025-01-03T17:56:28Z)
• On-the-fly Modulation for Balanced Multimodal Learning [53.616094855778954]
  Multimodal learning is expected to boost model performance by integrating information from different modalities. The widely-used joint training strategy leads to imbalanced and under-optimized uni-modal representations. We propose On-the-fly Prediction Modulation (OPM) and On-the-fly Gradient Modulation (OGM) strategies to modulate the optimization of each modality.
  arXiv  Detail & Related papers  (2024-10-15T13:15:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.