The paper introduces Uni-SafeBench, a safety benchmark designed to evaluate the holistic safety of Unified Multimodal Large Models (UMLMs) across six safety categories and seven task types. It also presents Uni-Judger, a framework that separates contextual safety from intrinsic safety to enable rigorous assessment. Evaluations reveal that current unified models do not consistently maintain the safety alignment of their underlying language models, and open‑source UMLMs perform significantly worse on safety than specialized multimodal models, especially in generation tasks.
By Zixiang Peng, Yongxiu Xu, Qin-Yi Zhang, Jiexun Shen, Yi-Fan Zhang, Hongbo Xu, Yubin Wang, Gaopeng Gou
The paper investigates the expressive power of multimodal contrastive learning architectures by treating them as parameterized families of joint density estimators. It shows that the classic two‑tower CLIP model is a universal approximator for two modalities, while a common extension that sums pairwise similarities fails to approximate arbitrary joint distributions when three or more modalities are involved, though it can match all pairwise conditionals. To address this limitation, the authors introduce Hadamard‑CLIP, which adds a single learned weight vector to restore universal approximation for any number of modalities while retaining CLIP’s efficient retrieval capabilities.
By Andrew Stuart, Florian Wolf
The paper introduces Role-Conditioned Sub-Token Routing (RoleSub), a method that compresses the value representations of retained tokens in Vision‑Language‑Action models. By partitioning each token’s value into orthogonal groups and routing them based on token features, latent roles, and language context, RoleSub can compress both visual and language representations without discarding tokens. Experiments on OpenVLA‑OFT‑7B show that, at matched visual‑KV budgets, RoleSub outperforms token‑only control in most settings and can reduce total KV to 9.2–11.3% of the original while maintaining strong control performance.
FedCoRe is a federated learning framework that addresses missing modalities in healthcare by learning representation- or logit-space corrections instead of generating synthetic data. In a MIMIC-derived respiratory deterioration task, the method uses paired examples where a modality is present during training but may be absent at deployment, allowing only those clients to update the completion module. Experiments show that FedCoRe can recover roughly half of the performance lost when ECG or CXR data are hidden, but the framework should only be deployed when paired examples and validation evidence confirm the modality’s presence.
Metasurfaces have revolutionized the development of photonic devices by enabling unprecedented precision in light manipulation. However, their design processes are often constrained by computationally...
The paper introduces a plug‑and‑play method that injects traffic‑element signals—such as traffic lights and road signs—into end‑to‑end autonomous driving models with minimal architectural changes. By augmenting several public datasets with comprehensive traffic‑element annotations, the authors evaluate this integration across diverse driving paradigms, consistently improving performance on nuScenes, NAVSIM‑v1, NAVSIM‑v2, and Bench2Drive. The approach achieves a new state‑of‑the‑art result on the challenging NAVSIM‑v2 benchmark, demonstrating the broad utility of traffic‑element awareness.
CFB-GBM v2.0 is an expanded longitudinal dataset of 264 glioblastoma patients, providing complete Gross Tumour Volume (GTV) delineations across all timepoints and derived volumetric RANO 2.0 response labels. The dataset includes brain masks, pre‑computed radiomic features, and WHO classification guidelines, all validated by radiation oncologists. It is publicly available on TCIA for use in computational methods for treatment response prediction and disease progression modeling.
The paper introduces TEAMS, a vision‑language Mamba snake framework that enhances deep snake instance segmentation. It adds a Spatiotemporal Snake Evolution Strategy to handle complex shapes, a Contour Morphology‑Aware Mamba to improve fine‑grained detail capture, and a Text‑prompted Collaborative Dual‑Head Snake to integrate textual cues and reduce detection errors. Experiments on five medical imaging datasets show TEAMS surpasses existing methods, achieving significant gains in mDice and mBF metrics.
MoE-ViE introduces a Mixture-of-Experts vision encoder that scales efficiently for image and video understanding, outperforming dense counterparts across various sizes. The study shows fine‑grained MoE topologies provide significant gains, and proposes an auxiliary‑loss‑free balancing variant and a specialized MoE kernel to reduce inference latency. With frame‑level distillation and a novel freezing mechanism, the largest MoE‑ViE model matches state‑of‑the‑art zero‑shot performance while being 1.7× larger and 76% faster, and it outperforms other encoders when paired with a language model on both image and video benchmarks.
CORAM introduces a new approach to merging fine‑tuned models by partitioning each target weight matrix into row slices and representing each slice with its singular value decomposition in the base‑model’s SVD frame. The method performs manifold averaging of task‑specific factors and applies an amplification coefficient to counteract contraction, with the coefficient’s scale estimated from update norms and its restoration strength chosen from expert update dispersion. Across multiple model families and scales, CORAM outperforms the prior OrthoMerge technique by up to 1.35 points and matches or exceeds the strongest weight‑space baselines.
The paper introduces Fair-ASR, a new evaluation protocol for black-box jailbreak attacks that uses shared target-call budgets to provide a fair comparison across methods. Re‑evaluating 11 attacks under this protocol shows that rankings shift significantly with different budgets, and that simple perturbations and templates remain competitive. The authors also present ReCode, a budget‑efficient attack that combines desensitization rewriting with low‑cost primitives, achieving 85% ASR on GPT‑5 with only 7.19 attacker calls per request under a 20‑call budget.
arXiv:2608. 15662v1 Announce Type: new Abstract: On modern e-commerce stores, customers consume ordered slates of heterogeneous product media, such as images, videos, and 3D renders, before making purchase decisions.
By Prasenjit Dey, Frank McIntyre, Arnab Sinha
arXiv:2608. 15690v1 Announce Type: cross Abstract: Text-to-audio-video (T2AV) generation models produce a video and its soundtrack from a textual description, but offer no control over whose voice speaks in the output.
By Ivan Mikheev, Viacheslav Vasilev, Anna Dmitrienko, Alexey Letunovskiy, Ivan Kirillov, Kirill Chernyshev, Denis Dimitrov
arXiv:2608. 15614v1 Announce Type: cross Abstract: The use of multimodal LLMs (MLLMs) for egocentric video understanding with wearable devices is constrained by the token budget.
By Matteo Stoiber, Niels Buus Lassen
arXiv:2608. 16889v1 Announce Type: cross Abstract: Long-horizon robot manipulation chains many contact-rich skills into one multi-stage task.
By Bingxin Xu, Yuzhang Shang, Emilio Ferrara
arXiv:2608. 15285v1 Announce Type: cross Abstract: Parameter-efficient fine-tuning (PEFT) is a natural way to adapt pretrained vision-language-action (VLA) policies, but most adapter designs apply temporally static updates throughout a control rollout, overlooking the phase-dependent nature of continuous-action manipulation.
By Yufei Guo, Yinan Wu, Haoran Duan, Guiguang Ding, Jungong Han
arXiv:2608. 15680v1 Announce Type: cross Abstract: Vision-language-action (VLA) models improve robotic manipulation but remain vulnerable to compounding errors, scene changes, and off-trajectory states.
By Yijie Xu, Haopeng Jin, Run Zhou, Shengbang Liu, Sixiang Chen, Hongyang Cheng, Sicheng Hu, Peterson Co, Jinwen Luo, Huajie Tan, Shanghang Zhang
arXiv:2608. 14763v1 Announce Type: cross Abstract: Assessment of ventriculomegaly (VM) on fetal brain ultrasound relies primarily on measuring lateral ventricular atrial width on standard planes, which is operator-dependent and may not fully reflect the overall ventricular enlargement.
By Yuhao Huang, Yuanji Zhang, Yuhuan Lu, Dong Ni, P. Ellen Grant, Davood Karimi
arXiv:2608. 16628v1 Announce Type: new Abstract: Modern Multimodal Retrieval-Augmented Generation (M-RAG) systems are fundamentally limited by the binary connectivity paradigm of traditional simple graphs, which fails to capture the intricate, high-order correlations among heterogeneous entities, such as the N-ary relationships between a visual chart, its scattered textual descriptions, and underlying numerical data.
By Shenao Chen, Yidan Xu, Xiangmin Han, Rundong Xue, Duanpo Wu, Yuhan Gao, Chenggang Yan, Yue Gao
arXiv:2608. 14741v1 Announce Type: cross Abstract: We introduce PolyComp, a procedurally generated and verified benchmark that stresses visual recognition and compositional spatial reasoning.
By Siddharth Patel