SpeakPay is a voice‑first digital wallet designed to make mobile payment apps in Nepal accessible to visually impaired users. The paper introduces NepFinSpeech‑403, a 403‑utterance Nepali financial voice command dataset, and demonstrates that fine‑tuning Whisper large‑v2 with LoRA reduces the Word Error Rate from 129.95% to 42.58% and improves Devanagari numeral recognition from 0.0% to 73.9%. Domain adaptation also boosts the Transaction Success Rate from 1.67% to 33.33%, with as few as 100 domain‑specific utterances halving the zero‑shot WER.
By Biraj Subedi
The paper introduces Temporal Context Routing (TCR), a method that aligns script timing with the shared temporal axis used for audio and video generation. By routing prompt guidance to specific temporal positions in both modalities, TCR dramatically improves shot boundary accuracy and dialogue timing on a set of test scripts, while preserving visual quality and audio‑visual sync. A user study confirms that participants prefer TCR across multiple evaluation dimensions.
By Yichen Liu, Quanwei Zhang, Haozhe Wang, Donghao Zhou, Xiaojie Li, Yang Shi, Jiaming Liu, Ruihua Huang, Yingtian Zou, Daquan Zhou
CrashDiffuser is a closed-loop VLM‑guided diffusion framework designed for fine‑grained safety‑critical traffic scenario generation. It separates semantic collision reasoning from trajectory synthesis via a hierarchical collision‑intent interface that specifies target contact regions (head, rear, or side). The system uses a vision‑language model to extract scene context and predict structured action tuples, which condition a diffusion model to produce executable adversarial trajectories, achieving high target‑collision and contact‑region control rates on WOMD‑derived scenarios.
By Shucheng Zhang, Yuang Zhang, Bingzhang Wang, Muhammad Monjurul Karim, Kehua Chen, Yinhai Wang
The paper investigates how Vision‑Language‑Action (VLA) models can generalise across different driving environments and camera setups. It introduces a multi‑dataset training strategy and an auxiliary objective called BEV‑Forcing, which injects bird‑eye‑view spatial information into the VLA backbone to improve both in‑distribution and out‑of‑distribution performance on a limited number of camera rigs. The authors observe that while BEV‑Forcing helps when training data is scarce, its advantage diminishes as the number of training embodiments grows, suggesting that scaling diversity may reduce the impact of such auxiliary tasks.
By Caio Azevedo, Stefano Sabatini, Sascha Hornauer, Fabien Moutarde
AlphaRAD introduces a grounded zero‑shot classification framework for chest radiology that leverages structured medical concepts extracted from reports and a novel α‑Corrected Binary Cross‑Entropy loss to reduce in‑batch noise. It also presents FLaS, a lightweight cross‑modal fusion module that factorizes VLPM representations into independent subspaces, improving spatial grounding without adding parameters. The method achieves state‑of‑the‑art performance on 16 classification benchmarks and sets new records on several grounding, phrase‑grounding, and segmentation datasets.
By Jianzhong You, Yuan Gao, Chris McIntosh
The study presents a multimodal BrainAGE framework that fuses structural T1-weighted MRI with DeepCBV maps—vascular information synthesized from non‑contrast MRI—to estimate brain age more accurately. Using two separate 3D convolutional neural networks, the combined model achieved a mean absolute error of 3.95 years on cognitively normal controls, outperforming single‑modality models. Saliency analyses showed MRI highlighted white matter and cortical atrophy, while DeepCBV emphasized vascular‑rich regions, and the integrated approach revealed stronger differentiation between stable and progressive MCI, indicating sensitivity to early vascular changes.
By Jordan Jomsky, Zongyu Li, Kay C. Igwe, Yiren Zhang, Max Lashley, Tal Nuriel, Andrew Laine, Scott A. Small, Jia Guo, for the Frontotemporal Lobar Degeneration Neuroimaging Initiative, for the Alzheimer's Disease Neuroimaging Initiative
The paper introduces OrthoRec, a multimodal recommender system that challenges the common "modality harmony" assumption by addressing conflicts between multimodal features and collaborative patterns. It employs Collaborative‑Guided Orthogonal Purification (CGOP) to separate useful multimodal signals from noisy orthogonal components, and a Topology‑Aware Routing Mixture‑of‑Experts (TAR‑MoE) to adaptively integrate purified modalities based on collaborative topology. Experiments on Amazon datasets demonstrate that OrthoRec outperforms recent baselines and remains robust to modality noise and item sparsity.
By Jialin Liu, Zhaorui Zhang, Ray C. C. Cheung
The paper investigates privacy risks in agentic AI systems that assemble sensitive data into a hidden context before responding. It introduces context‑inference attacks, a security game that evaluates how well attackers can recover this hidden context under varying levels of knowledge and indirect delivery. Experiments show that even with controls such as instructions not to disclose, logit suppression, and context dilution, agents can leak significant contextual information, achieving high success rates across multiple attack settings.
By Prince Jha, Samuele Poppi, Nils Lukas
MUDIDI is a two-stage framework designed to digitize multilingual dictionaries that are currently only available as scanned images. The first stage assesses character recognition and markup preservation, while the second stage segments dictionary entries and maps them into the SIL Multi-Dictionary Formatter schema. The authors also release a dataset of 30 annotated dictionaries and benchmark OCR, LLM, and VLM systems, finding that LLMs generally outperform others and that providing additional context improves digitization quality.
By David Setiawan, Temuulen Khishigsuren, Milind Agarwal, Pagnarith Pit, Aso Mahmudi, Ekaterina Vylomova
The paper investigates how much clinical text influences pixel‑level predictions in multimodal medical image segmentation. It shows that segmentation performance is largely insensitive to the choice of fusion module, but that the impact of text varies across datasets: removing text severely degrades performance on BUSI and BTMRI, while it has only a marginal effect on ISIC and Kvasir‑SEG. Using an Evidence Decoupling Decoder, the authors reveal that text mainly modulates global semantic context rather than spatial localization, and that the specific semantic components driving sensitivity differ by dataset.
By Ziquan Liu, Zhewei Zhu, Xuyang Shi
The paper investigates cross‑modal safety drift in multimodal large language models, where a harmless text query paired with a visual image can trigger harmful responses. Empirical analysis identifies unsafe response patterns and shows that visual cues receive limited attention, weakening refusal mechanisms. The authors introduce Safety‑Awareness Representation Transfer (SRT), a lightweight method that transfers safety signals from text processing to mitigate cross‑modal drift while maintaining model utility.
By Tianqi Xiao, Shiyao Cui, Minghao Zhang, Junxiao Yang, Renmiao Chen
The paper introduces Test-Time Logit Prompting (TLP), a lightweight framework that adapts vision-language models to missing-modality inputs without accessing source training data. TLP optimizes logit prompts using uncertainty-aware adjustments and modality-complete consistency regularization, thereby maintaining prediction confidence and semantic consistency. Experiments on various benchmarks show that TLP improves recognition performance by up to 8% while requiring only a few hundred tunable parameters and minimal test-time optimization steps.
By Taixi Chen, Nancy Guo
DiDrive introduces a risk‑aware hierarchical diffusion framework for offline reinforcement learning in autonomous driving. It combines a low‑level risk‑gated encoder with a high‑level contextual modulator to filter redundant state information, and a 3DICE policy optimization that reduces out‑of‑distribution overestimation and stabilizes gradients. On the CARLA benchmark, DiDrive outperforms baselines such as IQL, CQL, and Diffusion‑QL, achieving an 85% success rate and a 4295.68 average reward in dense traffic with 60 vehicles.
By Qisong Guo, Jingtang Chen, Zhilin Chen, Pei Xu, Mingjian Fu, Wenxi Liu, Yuanlong Yu
The paper examines six inference-time hallucination mitigation methods applied to three large vision-language models across four benchmarks, including MMStar. It finds that reducing hallucination rates often comes at the cost of lower informativeness—such as decreased object recall, visual coverage, and response detail—and that gains on hallucination benchmarks do not consistently translate to improved performance on fine-grained perception and reasoning tasks. The authors argue that current evaluation protocols may overstate progress by favoring conservative generation, and propose that hallucination mitigation should be assessed as a trade-off among faithfulness, informativeness, and overall capability.
By Mehrdad Fazli, Sina Mansouri, Mohit Marvania, Ziwei Zhu
SafeEvolve is an experience-driven framework that co‑evolves a harness and policy to align large‑language‑model agents with safety goals. It uses completed on‑policy trajectories to update safety prompts and hierarchical skills, then applies a two‑stage SFT‑RL training loop that bootstraps the policy with the evolved harness and refines it through verifier‑augmented rewards. Experiments on agentic safety benchmarks show that SafeEvolve improves the safety‑utility tradeoff, achieving a three‑fold reduction in ASR on AgentDojo for Qwen3.5‑4B while increasing benign utility from 59.79% to 61.86%.
By Qinghua Mao, Wanying Qu, Dadi Guo, Leitao Yuan, Qingyu Liu, Yu Li, Guanxu Chen, Yanwei Fu, Xi Lin, Xia Hu, Dongrui Liu
Multimodal Language Models as Text-to-Image Model Evaluators presents MT2IE, a framework where a multimodal large language model generates evaluation prompts and scores images, achieving higher correlation with human judgment than prior metrics. MT2IE recovers official T2I model rankings using only 20 prompts—far fewer than traditional benchmarks—and adapts prompts to each model’s performance, maintaining informative scoring ranges. The approach demonstrates that dynamic, interactive evaluation can replace static benchmarks as T2I models improve.
By Jiahui Chen, Candace Ross, Reyhane Askari-Hemmat, Koustuv Sinha, Melissa Hall, Amy Zhang, Michal Drozdzal, Adriana Romero-Soriano
FemWear is a parameter‑efficient wearable foundation model specifically tailored for women's health. It repurposes a pretrained multimodal wearable backbone by training only 239,236 encoder parameters—just 1.11% of the original 21.54M—using low‑rank residual adapters and causal task‑family heads to create a shared longitudinal representation for menstrual, symptom, affective, sleep/recovery, autonomic, activity, and pregnancy outcomes. Evaluations across six cohorts and 63 metrics show improvements in cycle‑phase macro‑F1 and reductions in mean absolute error for cramps, mood symptoms, and sleep problems, while maintaining the OpenMHC ability‑retention benchmark.
By Yifan Wang, Chenzhong Li
The paper introduces TIC‑Bench, a new benchmark for evaluating multimodal large language models on deeply interleaved text‑image contexts. It covers logical, temporal, and spatial association tasks, totaling 2,280 questions across eight specific types. The authors benchmarked ten state‑of‑the‑art MLLMs, finding a significant performance gap versus human experts and highlighting persistent challenges in integrating evidence across interleaved visual and textual inputs.
By Zihao Wang, Xi Xiang, Yuwen Sun, Yingyu Li, Yabo Zhang, Yihan Zeng, Fan Li, Wangmeng Zuo
FairLens is a benchmark and evaluation framework that measures fairness and validity of vision‑language models (VLMs) in high‑stakes domains such as hiring, legal, and healthcare. It uses over 100,000 face‑image and question pairs covering gender, race, and age, and assesses responses through demographic parity, soundness, demographic association, and bias in free‑text generation. The study finds that VLMs often make unwarranted inferences from faces rather than abstaining, especially in legal and healthcare contexts, and that small parity gaps can still hide unsafe treatment across groups.
By Vahid Reza Khazaie, Ahmed Y. Radwan, Shaina Raza
CACTUS is a new backdoor attack for decentralized federated learning that uses mask‑guided, modality‑specific operators to convert label‑consistent semantic pairs into target‑directed representation shifts. By isolating trigger effects and applying them counterfactually to clean embeddings before peer aggregation, CACTUS can propagate backdoors across repeated aggregation rounds. Experiments on speech, text, tabular, and image tasks show that with 30% malicious nodes, CACTUS achieves a mean attack success rate of 51.2% on Speech Commands and the highest mean ASR among evaluated attacks on three of four modalities, with success varying by network topology and malicious‑node ratio.
By Chao Feng, Burkhard Stiller