ARGenSeg introduces an autoregressive generation-based approach for image segmentation that integrates seamlessly with multimodal large language models (MLLMs). Unlike prior methods that use boundary points or dedicated segmentation heads, ARGenSeg generates dense masks directly through visual token output and detokenization via a universal VQ‑VAE, enabling fine‑grained pixel‑level perception. The framework employs a next‑scale‑prediction strategy to parallelize token generation, resulting in faster inference while outperforming state‑of‑the‑art segmentation models on multiple datasets.
By Xiaolong Wang, Lixiang Ru, Ziyuan Huang, Kaixiang Ji, Dandan Zheng, Jingdong Chen, Jun Zhou
RAVE (Re-Allocating Visual Attention) is a lightweight pair‑gating mechanism that adds a learned query‑key bias to pre‑softmax attention scores over visual keys, derived from pre‑RoPE query and key features. It requires no architectural changes to the backbone and can be trained end‑to‑end with the rest of the model. Across multiple multimodal benchmarks, RAVE improves standard attention by an average of 3 points, especially on perception‑intensive tasks such as multilingual OCR, chart understanding, document VQA, and scene text VQA.
By Xi Leng, Xinhong Ma, Ziqiang Dong, Feng Zhang, Xiaoying Tang, Yang Yang, Guanjun Jiang
The study investigates why language models exhibit systematic performance gaps across English dialects, a phenomenon termed the "dialect tax." Using parallel dialect corpora that preserve meaning while altering surface form, the authors confirm that models treat Standard American English and dialectal texts as semantically equivalent, yet find representational disparities that persist through tokenization, pre‑training, post‑training, and inference. Even a character‑level tokenizer does not eliminate input/output asymmetries or accuracy gaps, and dialect pairs produce more divergent gradient updates than unrelated Standard texts, indicating that dialectal content is harder for models to learn.
By Elle
ClueWeaver is a dual-agent framework designed to enable compact, locally deployable language models to answer questions about long literary narratives. The Finder agent retrieves passages that contain answer-critical clues, while the Interpreter agent derives the answer from those passages, generates rationales with paragraph-ID citations, and performs self-calibration for high-risk questions. Both agents are trained with reward-guided reinforcement learning to prioritize evidence retention, correctness, grounding, and concise explanations, resulting in improved performance and inspectability over end-to-end prompting.
By Jihao Zhu, Zhiwei Yang, Wenxiao Zhang, Junqian Zhao, Qi You, Fangqi Wang, Zheyuan Deng, Hanzhe Yang, Yu Liu, Jin B. Hong
The paper presents a modality‑routed 3D cardiac segmentation pipeline that combines TotalSegmentator‑initialized nnU‑Netv2 models with site‑characterized, label‑preserving appearance augmentation. By analyzing measurable image properties across sites, the authors design a bias‑field plus Bezier augmentation strategy that smooths spatial intensity perturbations and remaps intensities nonlinearly, followed by class‑wise largest‑connected‑component cleanup. On held‑out validation splits, this approach raises CT mean Dice from 0.8350 to 0.9135 and MRI mean Dice from 0.7695 to 0.7830 while reducing HD95, demonstrating improved cross‑site robustness in limited‑data whole‑heart segmentation.
By Tanish Mudaliar, Justin Li, Daniel Lin, Julianna Vo, Kaitao Liao, Xin Wang, Shu Hu
RSFusionDet introduces a new RGB‑Sonar multimodal object detection dataset (RSFusion) and benchmark metrics for underwater imaging. The proposed detector uses a Cross‑Attention Fusion module to align RGB and sonar features and an Object Matching Head with loss to identify identical objects across modalities. On the RSFusion dataset, RSFusionDet achieves 76.4/48.6 AP for RGB/sonar detection and 83.4 F1‑Score for cross‑modal matching, outperforming existing models and improving over the DINO baseline by 0.7/1.4 AP.
By Zhuoyan Liu, Yihan Wang, Bo Wang, Bing Wang, Ye Li
The paper introduces grounded glossary generation, a structured NLP task that asks models to recover semantically meaningful Sanskrit phrases and provide translation‑grounded meanings from a sloka‑translation pair, mirroring the traditional patha commentary practice. A benchmark of 31,316 sloka‑translation‑glossary triples from the Valmiki Ramayana and Srimad Bhagavatam is built, evaluated with Jaccard for phrase recovery and Meaning Faithfulness for semantic consistency. Experiments with Gemma‑3n‑E4B, Gemma‑3‑12B, Phi‑4, and Qwen3.5‑9B show that instruction fine‑tuning outperforms prompting, and explicit segmentation further improves results, though over‑segmentation of sandhi and samasa compounds remains the main error source, highlighting morphological modeling as a key bottleneck.
By Manoj Balaji Jagadeeshan, Sai Pragnaan Marala, Pawan Goyal
The paper introduces SAME (Stabilized Mixture-of-Experts) to address challenges in Multimodal Continual Instruction Tuning (MCIT) for large language models. SAME mitigates router drift by decomposing routing dynamics into orthogonal subspaces and updating only task-relevant directions, while preventing expert drift through curvature‑aware scaling that uses historical input covariance without rehearsal. The method also employs adaptive expert activation to freeze selected experts during training, reducing redundant computation and cross‑task interference, and demonstrates state‑of‑the‑art performance on a new long‑task‑sequence benchmark.
By Zhen-Hao Xie, Jun-Tao Tang, Yu-Cheng Shi, Han-Jia Ye, De-Chuan Zhan, Da-Wei Zhou
The paper introduces a label‑free method called AURCC for selecting the best foundational model for medical image classification when the target domain lacks labels. AURCC uses a pseudo‑label discrepancy computed by the SUDO framework to score models without fine‑tuning. Experiments on chest X‑ray data across three inter‑hospital shifts show that AURCC closely matches the true model ranking, outperforming simple source‑accuracy baselines especially when source data are limited.
By Juan I\~naki Larrea, Lucas Mansilla, Enzo Ferrante
The paper introduces Socialized Detector Learning (SDL), a framework for improving heterogeneous, category-specialized object detectors by exchanging knowledge within a detector society. It proposes Trajectory-Guided and Reciprocal Distillation (TGRD), which estimates transfer difficulty, constructs a carrier trajectory, and performs progressive consolidation and reciprocal transfer. Experiments on MS COCO show that carriers outperform simultaneous aggregation by 2.6 AP, and reciprocal detectors achieve high AP on previously unsupported categories while staying close to expert performance.
By Weihao Li, Yunqi Zhu, Zhihe Fan, Ruipu Zhao, Boan Tao, Xinjie Yao, Yan Fan, Pengfei Zhu
MIMONet is a saliency detection model that uses multi‑scale inputs and outputs to better handle objects of varying sizes. It processes three differently sized images through separate encoder branches that exchange information, allowing each branch to learn size‑variation knowledge from the others. A Multi‑scale Perception module further refines features, and a Joint Saliency Loss ensures consistent, well‑preserved boundaries across the multiple saliency maps produced.
By Zhaojian Yao, Wei Gao, Tiesong Zhao, Hui Yuan, Sam Kwong
The paper introduces a structure‑aware merging pipeline that consolidates per‑pixel 3D Gaussian primitives from any feed‑forward reconstruction method into a compact, content‑adaptive Gaussian set. By grouping spatially coherent Gaussians with adaptive superpixel segmentation guided by a saliency map, compressing clusters via a learned encoder, and merging representations across views using geometric overlap and feature similarity, the method reduces the number of Gaussians to about one‑twentieth of the original while preserving visual quality. A level‑of‑detail decoder allows controllable resolution, and the pipeline operates as a backbone‑agnostic post‑processing module, improving robustness and rendering efficiency.
By Tim-Felix Fassch, Jochen Kall, Cyrill Stachniss
LUTSeg is a new longitudinal dataset for chronic ulcer tissue segmentation, containing 141 images from 39 patients with wound masks and five tissue categories annotated by five expert clinicians. The dataset includes a multi‑expert gold‑standard subset for inter‑rater agreement analysis. The authors also propose TiSage, a semi‑supervised segmentation framework that uses multi‑scale semantic priors from a frozen medical vision‑language model within a teacher‑student architecture, and demonstrate its effectiveness on LUTSeg and DFUTissue.
By Karen Sanchez, Carlos Hinojosa, Albert A. \'Avila, Andrea C. Riano-Rojas, Diego H. Romero, Jenny C. P\'aez, Martina Llin\'as, Bernard Ghanem
The study evaluated a pragmatic deep‑learning approach for segmenting acute ischemic stroke lesions on diffusion‑weighted MRI. Using a self‑configured nnU‑Net trained on 1,744 cases and tested on 436, the baseline model achieved a median Dice similarity coefficient of 0.84, outperforming the DeepISLES ensemble, especially for smaller infarcts. The approach required minimal preprocessing and fast inference, suggesting it could streamline clinical stroke imaging workflows.
By Atle Bj{\o}rnerud, Till Schellhorn, Thor H. Skatt{\o}r, Terje Nome, Jon Andr\'e Ottesen, Anne Hege Aamodt, Bradley J MacIntosh
PlanSightRAG is a visual-first multimodal retrieval‑augmented generation system designed to automate question answering and compliance checking of civil standard plans. It processes plan imagery directly, using a ColNomic‑3B multi‑vector retrieval engine, an agentic Planner‑Retriever‑Auditor‑Synthesizer, and MaxSim heatmaps to provide an evidence trail. The system achieves high recall on a new 4,056‑pair benchmark from five state DOTs, and demonstrates near‑perfect verdict accuracy on synthetic compliance drawings when a rule threshold is supplied, outperforming OCR‑based baselines.
By Nabaraj Subedi, Shuvo Dip Datta, Ahmed Abdelaty, Shivanand Venkanna Sheshappanavar
The study investigates how confidence intervals (CIs) behave in medical imaging AI by analyzing 24 segmentation and classification tasks with 19 models per task, various metrics, aggregation strategies, and CI methods. It finds that required sample sizes for reliable CIs vary widely, CI behavior depends on performance metrics, aggregation strategy, and problem type, and that different CI methods differ in reliability and precision. The authors provide a decision tree to guide researchers in selecting appropriate CI methods, aiming to support future consensus guidelines on reporting performance uncertainty.
By Pascaline Andr\'e (Sorbonne Universit\'e, Institut du Cerveau - Paris Brain Institute - ICM, CNRS, Inria, Inserm, AP-HP, H\^opital de la Piti\'e-Salp\^etri\`ere, Paris, France), Charles Heitz (Sorbonne Universit\'e, Institut du Cerveau - Paris Brain Institute - ICM, CNRS, Inria, Inserm, AP-HP, H\^opital de la Piti\'e-Salp\^etri\`ere, Paris, France), Evangelia Christodoulou (German Cancer Research Center), Annika Reinke (German Cancer Research Center), Carole H. Sudre (Unit for Lifelong Health and Ageing at UCL, Department of Population Science and Experimental Medicine and Hawkes InstituteCentre for Medical Image Computing, Department of Computer Science, University College London, UK), Michela Antonelli (School of Biomedical Engineering and Imaging Science, King's College London, UK), Patrick Godau (German Cancer Research Center), M. Jorge Cardoso (School of Biomedical Engineering and Imaging Science, King's College London, UK), Antoine Gilson (Sorbonne Universit\'e, Institut du Cerveau - Paris Brain Institute - ICM, CNRS, Inria, Inserm, AP-HP, H\^opital de la Piti\'e-Salp\^etri\`ere, Paris, France), Sophie Tezenas du Montcel (Sorbonne Universit\'e, Institut du Cerveau - Paris Brain Institute - ICM, CNRS, Inria, Inserm, AP-HP, H\^opital de la Piti\'e-Salp\^etri\`ere, Paris, France), Ga\"el Varoquaux (SODA project team, Inria Saclay-\^Ile-de-France, France), Lena Maier-Hein (German Cancer Research Center), Olivier Colliot (Sorbonne Universit\'e, Institut du Cerveau - Paris Brain Institute - ICM, CNRS, Inria, Inserm, AP-HP, H\^opital de la Piti\'e-Salp\^etri\`ere, Paris, France)
The paper presents 3DGAA, a fabrication-first framework that generates view-consistent, geometry-preserving adversarial wraps for vehicles using 3D Gaussian splatting optimization. It ensures consistency across viewpoints, illumination, and occlusion while limiting changes to vehicle geometry, producing realistic print-only textures that significantly reduce detection confidence and average precision in simulations and physical tests. Ablation and efficiency studies analyze the impact of physical filtering, augmentation, and shape-consistency regularization, and the method demonstrates robustness against common preprocessing defenses and cross-detector transferability.
By Yixun Zhang, Lizhi Wang, Junjun Zhao, Wending Zhao, Feng Zhou, Yonghao Dang, Jianqin Yin
TDFNet introduces a Tri-projection Deformable Fusion Network that uses equirectangular, cube map, and tangent projections to mitigate geometric distortions in panoramic salient object detection. It incorporates a cross-projection deformable attention module for geometry-aware sampling and a latitude-guided fusion module that balances ERP and CMP features using spherical latitude priors. The network’s three-branch encoding preserves global continuity, local detail, and boundary precision, improving detection performance over existing projection-based methods.
By Qiangqiang Zhou, Jiacong Yu, Jiawei Xu, Yong Chen, Xin Huang, Ping Li
The paper introduces MSM‑Seg, a dual‑memory segmentation framework for 3D multi‑modal brain tumor segmentation. It combines a modality‑and‑slice memory attention module to capture cross‑modal and spatial‑slice dependencies, a multi‑scale category‑agnostic prompt encoder for whole‑tumor guidance, and a modality‑adaptive fusion decoder to integrate complementary decoding information. Experiments on various MRI datasets show that MSM‑Seg surpasses state‑of‑the‑art methods for metastases and glioma tumor segmentation.
By Yuxiang Luo, Qing Xu, Hai Huang, Yuqi Ouyang, Xiangjian He, Zhen Chen, Wenting Duan, Jiebo Luo
Multi-norm adversarial defense aims to protect neural networks against perturbations defined by different norm constraints, but existing methods typically optimize competing robustness objectives with...