Vision-language models, speech and cross-modal systems that read, look and listen in the same forward pass.
arXiv:2608.08273v2 Announce Type: replace-cross Abstract: Vision-based embodied agents executing multi-step natural language instructions require feedback mechanisms that assess task progress over co...
CommerceVibe is a system that generates e‑commerce creatives by synthesizing executable HTML/CSS code conditioned on product images, design requirements, and product information. It uses dual‑feedback reinforcement learning, combining rule‑based checks for text readability, product visibility, and layout validity with visual feedback from a vision‑language model that evaluates perceptual and commercial aspects. After fine‑tuning a large language model on 28,000 examples and applying dual‑feedback reinforcement learning, CommerceVibe achieves a weighted score of 94.0/100 on a 1,300‑case benchmark, outperforming both its SFT‑only counterpart and external models, and is validated by expert blind evaluations.
The paper introduces a Mixture of Multicenter Experts (MoME) framework that leverages diverse clinical strategies to reduce bias in medical AI without sharing data across institutions. MoME integrates specialized expertise from multiple centers, improving generalizability and adaptability of a multimodal target volume delineation model for prostate cancer radiotherapy. The model, trained with few-shot imaging and clinical notes, outperformed baselines, especially in high inter‑center variability or limited data scenarios, and allows local customization without cross‑institutional data exchange.
The paper introduces Dual-Stream Semantic Guidance (DSSG), a framework for Source‑Fully‑Free Domain Adaptation of Vision‑Language Models that mitigates dual semantic drift through a caption stream and a class‑anchor stream. It adds a Dynamic Cross‑Modal Knowledge Distillation module and a Prototype Anchor Calibration extension (DSSG‑PAC) to reduce computation while maintaining performance. Experiments show DSSG outperforms state‑of‑the‑art methods and DSSG‑PAC cuts adaptation time by 18.9% with minimal loss in accuracy.
The paper introduces VLAct, a Vision‑Language‑Action model that focuses on representation‑centric continued pre‑training rather than merely scaling robot data. VLAct is trained on diverse, multi‑embodiment robot data and preserves a broad VLM prior while encouraging shared action semantics across embodiments. Experiments across simulation, real‑world, and unseen‑embodiment settings show that VLAct consistently outperforms existing industrial VLA systems, achieving high success rates with only a modest compute budget and open‑source data.
The paper introduces Meta-Adaptive Multimodal Jailbreaking (MAMJ), a method that jointly optimizes an attack strategy prompt and attacker weights to generate more effective jailbreaks against vision‑language models. Using an LLM‑based critique to refine the strategy and group‑level success‑rate rewards to update the weights, MAMJ achieves high attack success rates on MM‑SafetyBench, outperforming existing baselines by up to 24.1 percentage points. The learned attacker also transfers to unseen models and remains robust against typical defenses, highlighting a systemic vulnerability in current VLMs.
Token-Budget Distillation (TBD) is a parameter‑efficient fine‑tuning framework that adapts video vision‑language models to a fixed token budget. It freezes the pretrained backbone, updates only LoRA adapters, and incorporates FlashVID visual token compression. TBD uses a dual‑path teacher‑student design with full‑token supervision and compressed student optimization, enabling the student to recover full‑token semantics while remaining efficient under aggressive token reduction.
iOSPointMapper is a mobile app that performs real‑time, privacy‑conscious sidewalk mapping using on‑device semantic segmentation, LiDAR depth estimation, and fused GPS/IMU data on recent iPhones and iPads. It detects and localizes sidewalk‑relevant features such as traffic signs, traffic lights, and poles, and includes a user‑guided annotation interface for validating outputs before submission. The anonymized data is transmitted to the Transportation Data Exchange Initiative (TDEI), where it integrates with broader multimodal transportation datasets, and evaluations show the app’s potential for enhanced pedestrian mapping.
The paper proposes GeBDA, a method that treats Building Damage Assessment as a text-based sequence prediction task using a Vision‑Language Model. Instead of specialized architectures, it uses an autoregressive VLM to output a variable‑length list of bounding boxes with coordinates and damage labels. The authors demonstrate promising results on bi‑temporal satellite images using the open Gemma model and a tailored text prompt.
The paper introduces QK Product Steering, a data‑free, training‑free method that edits the query‑key product in vision‑language models to reduce object hallucination. By suppressing a few dominant singular modes in selected middle layers and mapping the edited product back to query weights, the approach lowers hallucination rates without affecting inference cost. Experiments on three GQA‑based VLMs show a 4.0% average reduction in CHAIR$_s$, with the effect localized to symmetric mutual‑attention channels.
OmniFusion is an end‑to‑end multilingual multimodal translation system that fuses a pretrained multimodal foundation model (Omni 2.5‑7B) with a translation large language model (SeedX PPO‑7B). By connecting hidden states from multiple layers of the multimodal model to the translation LLM, OmniFusion can translate speech, speech‑and‑image, and text‑and‑image inputs while reducing simultaneous speech‑translation latency by about one second compared to cascaded pipelines. The approach improves overall translation quality by leveraging both audio and visual context.
The paper presents a multimodal 3D convolutional neural network that classifies Alzheimer’s disease using raw OASIS 1 MRI volumes. It fuses structural T1 images with gray matter, white matter, and cerebrospinal fluid probability maps to capture complementary neuroanatomical information. Evaluated with 5‑fold subject‑level cross‑validation, the model achieves a mean accuracy of 72.34 % and an ROC AUC of 0.7781, with GradCAM visualizations highlighting anatomically relevant regions such as the medial temporal lobe and ventricles.
SemEnrich introduces a self‑supervised method to enrich radiology reports by clustering sentences semantically and adding positive or neutral observations from different clusters. The enriched data consistently improves supervised fine‑tuning across multiple vision‑language metrics, with gains ranging from 3% to 7.5% on COMET, Bert, Sentence Bleu, CheXbert‑F1, and RadGraph‑F1. The authors also demonstrate that incorporating cluster information into the reward design for GRPO training yields additional performance boosts.
The paper introduces Salience-LLaVA, a vision‑language model that prioritizes scene elements based on their importance for low‑vision users. It presents three new salience‑aware datasets—Salience COCO, Salience Flickr, and Salience VizWiz—annotated with object‑level salience verified by low‑vision participants. The authors also propose the SCMI metric to evaluate caption ordering accuracy and demonstrate the system’s practicality by deploying it on assistive glasses.
The paper introduces TransMod, a unified framework for forecasting urban mobility demand across multiple transportation modes. It creates a shared zone-level spatial representation to align systems with different spatial granularities, reducing structural mismatch and distributional shift. TransMod then learns transferable spatio-temporal patterns from data-rich source modes and adapts them to data-scarce target modes, improving forecasting performance when target data is limited.
The paper introduces Short‑Films 20K (SF20K), a large publicly available movie dataset comprising 20,143 amateur films totaling 3,582 hours, with an average length of 12 minutes per film. Accompanying the dataset is SF20K‑Test, a manual open‑ended question‑answering benchmark featuring 95 movies and 979 question‑answer pairs. Analysis of the benchmark shows limited data leakage, highlights the necessity of long‑term reasoning, and demonstrates that instruction tuning on the large‑scale dataset significantly boosts vision‑language model performance.
Abstract4D is the largest dataset of abstract paintings, containing over 120,000 images with rich metadata and multi‑dimensional prompts that capture perceptual attributes such as form, color, texture, and composition. The dataset is annotated via a hybrid human–VLM pipeline to ensure quality and consistency. Using Abstract4D, the authors analyze the semantic structure of abstract art through large‑scale embedding visualization and establish benchmark tasks for classification, cross‑modal retrieval, and text‑to‑image generation to evaluate AI models’ perception and reproduction of abstract visual language.
uScenes is a new multimodal dataset for underwater robot perception that provides synchronized 3D multibeam sonar point clouds and RGB imagery. It comprises 110 scenes with 95,834 observations, totaling 277.6 minutes of data collected during multiple field sessions. The dataset aims to support research in underwater sensor fusion, cross‑modal representation learning, and 3D scene understanding.
The paper introduces Report Supervision (R‑Super), a framework that uses radiology reports to supervise tumor segmentation models. By incorporating loss functions that align segmentation outputs with report‑derived tumor counts, sizes, and locations, R‑Super improves detection and segmentation performance. Experiments on kidney and pancreatic tumors show up to a 15% increase in F1‑Score and DSC compared to mask‑only training, outperforming methods like CLIP and multi‑task learning.
arXiv:2608.27549v1 Announce Type: new Abstract: Physical understanding and reasoning depend on forming compact and generalizable representations of the world. While modern vision-language models can...