The paper investigates whether layer-wise visual‑text similarity in multimodal large language models (MLLMs) truly reflects content‑level cross‑modal interaction. By injecting Gaussian noise into the visual stream of 13 MLLMs, the authors show that task accuracy drops sharply while traditional scalar alignment metrics (CKA, SVCCA, MIR, principal‑angle cosine) fail to distinguish corrupted from clean inputs, a phenomenon they term the alignment illusion. They propose the principal‑angle gap (PA gap) as a more reliable geometric diagnostic that correlates better with task performance and reveals when internal geometry diverges from accuracy.
By Hong-Han Wang, Yuntao Wang, Hu Ding
arXiv:2602. 07026v3 Announce Type: replace-cross Abstract: Despite the success of multimodal contrastive learning in aligning visual and linguistic representations, a persistent geometric anomaly, the Modality Gap, remains: embeddings of distinct modalities expressing identical semantics occupy systematically offset regions.
By Xiaomin Yu, Yi Xin, Yuhui Zhang, Wenjie Zhang, Chonghan Liu, Hanzhen Zhao, Chen Liu, Xiaoxing Hu, Ziyue Qiao, Hao Tang, Xiaobin Hu, Chengwei Qin, Hui Xiong, Yu Qiao, Shuicheng Yan
arXiv:2608. 19598v1 Announce Type: cross Abstract: Direct Preference Optimization (DPO) has emerged as an effective approach for aligning large language models (LLMs) with human preferences.
By Jiawei Feng, Jiancan Wu, Xingyu Zhu, Junkang Wu, Xiang Wang, Xiangnan He
The paper investigates how visual and textual information are fused in Multimodal Large Language Models (MLLMs). By analyzing concatenation and native multimodal architectures through alignment decoupling, attention routing, entropy, intrinsic dimensionality, and causal interventions, the authors uncover two distinct fusion pathways: concatenation models use a text‑first, vision‑later strategy, while native models integrate vision and text earlier and reorganize feature spaces. The study also employs visual CKA to test the Platonic Representation Hypothesis, offering a mechanistic view of multimodal fusion and informing architecture‑aware diagnostics.
By Hebao Zhu, Dongxia Wu
arXiv:2606. 26387v1 Announce Type: cross Abstract: Multimodal large language models (MLLMs) extend large language models (LLMs) with visual perception, enabling joint reasoning over images and text.
By Xi Xiao, Chen Liu, Chih-Ting Liao, Yunbei Zhang, Qizhen Lan, Yuxiang Wei, Lin Zhao, Janet Wang, Jianyang Gu, Muchao Ye, Tianyang Wang, Hao Xu
arXiv:2605. 18160v2 Announce Type: replace-cross Abstract: In recent years, multimodal large language models (MLLMs) have achieved remarkable progress, primarily attributed to effective paradigms for integrating visual and textual information.
By Xinpeng Dong, Min Zhang, Kairong Han, Xu Tan, Fei Wu, Kun Kuang
The paper introduces GAP, a Granular Alignment Paradigm that improves visual latent reasoning in multimodal large language models by aligning latent representations at data, feature, and model levels. GAP uses auxiliary visual supervision, a PCA‑aligned latent head, and selective latent supervision to address feature‑space mismatches that destabilize prior methods. On Qwen2.5‑VL 7B, GAP outperforms other latent methods, enhancing both perception and reasoning scores while demonstrating that generated latents carry task‑relevant visual signals.
By Yanting Miao, Yutao Sun, Dexin Wang, Mengyu Zhou, Pascal Poupart, Lei Lv, Qi Zhao, Li Wang, Hao Li, Xiaoxi Jiang, Guanjun Jiang
The paper investigates how the modality gap— the separation between image and text representations in contrastive vision‑language models—affects different downstream tasks. By showing that a single dominant direction accounts for most of the image‑text mean separation, the authors explain why reducing or removing this gap can improve zero‑shot classification, degrade retrieval, or restore performance depending on the task. The study provides a geometric framework that clarifies when and why gap interventions should be applied in vision‑language systems.
By Aditya Sharma, Divya Saxena
arXiv:2308. 06035v4 Announce Type: replace Abstract: Humans routinely draw on visual context to predict upcoming words.
By Viktor Kewenig, Andrew Lampinen, Samuel A. Nastase, Christopher Edwards, Quitterie Lacome D'Elascombe, Akilles Rechardt, Jeremy I Skipper, Gabriella Vigliocco
DRAMA is a training‑free acceleration framework for multimodal diffusion large language models that dissects attention redundancy by probing attention tensors along spatial and temporal axes. It identifies two key redundancy patterns: a sparse subset of visual tokens consistently receives most attention, and attention distributions stabilize quickly after an initial phase. Leveraging these insights, DRAMA introduces Spatial‑Redundancy‑Aware Visual Token Pruning and Sufficiency‑Guided Adaptive Stopping, achieving an average 18.67× inference speedup across six vision‑language benchmarks while preserving competitive accuracy.
By Yi-Hsin Hung, Fangfu Liu, Shao-Yuan Lo, Chu-Song Chen
arXiv:2605. 08245v4 Announce Type: replace-cross Abstract: Vision-Language Models (VLMs) increasingly power high-stakes applications, from medical imaging to autonomous systems, yet they routinely hallucinate, confidently describing content not present in the input.
By Harshvardhan Saini, Samyak Jha, Yiming Tang, Dianbo Liu
arXiv:2604. 22823v2 Announce Type: replace-cross Abstract: Multimodal Large Language Models (MLLMs) rely on multimodal pre-training over diverse data sources, where different datasets often induce complementary cross-modal alignment capabilities.
By Zibo Shao, Baochen Xiong, Xiaoshan Yang, Yaguang Song, Qimeng Zhang, Haifeng Chen, Changsheng Xu