The paper introduces BROT, a two‑step approach for estimating optimal transport maps. First, it computes the unregularized OT plan, then fits a deep neural network to the resulting barycentric targets using least‑squares regression. The authors prove that, under standard regularity conditions, BROT achieves the minimax convergence rate when the true OT map is Lipschitz, and demonstrate its effectiveness on synthetic data, images, and downstream tasks such as single‑cell perturbation prediction and unsupervised domain adaptation.
By Kunwoong Kim, Insung Kong, Yongdai Kim
Osprey is a target‑agnostic pre‑training method that bootstraps draft models for speculative decoding from existing small language models. By pruning to a shallow backbone, restoring language‑modeling capability with next‑token pretraining, and adapting via vocabulary alignment and distillation, Osprey reduces per‑target work to a lightweight adaptation step. Experiments show that a single Osprey backbone improves mean acceptance length by up to 22.7% and increases tokens per second by 17.5% across several large target models, especially on out‑of‑domain and multilingual data.
By Fengxiang Bie, Yuqing Jian, Yifan Yu, Zhongzhu Zhou, Zelei Shao, Ben Athiwaratkun, Shuaiwen Leon Song, Chenfeng Xu, Xiaoxia Wu, Tianyi Zhang
The paper introduces REASONS, a benchmark of 12,723 sentence-level citation instances across 12 arXiv subject categories, to evaluate scientific citation attribution under different evidence conditions. It proposes a dual-metric framework—Abstention Rate (AR) and Hallucination Rate (HR)—to balance reliability and responsiveness. Experiments with proprietary and open-source LLMs across various prompting and retrieval settings show that advanced Retrieval-Augmented Generation (RAG) reduces hallucinations but increases abstention, while adversarial metadata can push hallucination rates above 85%. Human evaluation confirms a high ratio of factual hallucinations to acceptable paraphrases, underscoring the need for systems that can appropriately abstain under uncertainty.
By Deepa Tilwani, Yash Saxena, Seyedali Mohammadi, Ankur Padia, Edward Raff, Amit Sheth, Srinivasan Parthasarathy, Manas Gaur
The paper introduces a framework that learns the kernel used in kernel methods through alignment, leveraging the Collaborative Learning and Inference (CLaI) approach. It demonstrates that CLaI can be interpreted as a kernel alignment process and that its inference stage is equivalent to kernel Bayes classification with Parzen-window density estimation. By replacing cosine similarity with a learned Mahalanobis distance, the authors extend CLaI to multiclass classification, achieving higher accuracy, faster convergence, and lower calibration error on datasets such as CIFAR-10, PathMNIST, and SleepEDF, while also showing connections to Gaussian processes and competitive calibration in sepsis prediction.
By Hollan Haule, Alfredo Gonzalez-Sulser, Javier Escudero
The paper introduces the Local Gradient Neural Operator (LGNO), a lightweight and interpretable neural operator designed for field temporal evolution prediction and source identification in mechanical problems. LGNO leverages nonlinear gradient discretization priors and multilayer perceptron convolutional layers to learn translation‑invariant local kernels resembling discrete stencils, with a zero‑consistent stencil factorization that separates coefficient learning from field reconstruction. Experiments on a range of PDE benchmarks—including linear, nonlinear, static, dynamic, low‑ and high‑dimensional cases—demonstrate that LGNO achieves comparable accuracy to global neural operators while using fewer parameters and maintaining rollout stability across diffusion, flow, and quantum phenomena.
By Baiming Zhang, Jinsong Tang, Ying Xu, Lihua Chen, Shiying Xiong
Shortcut learning occurs when classifiers rely on spurious correlations instead of true diagnostic features. The authors argue that this is essentially a calibration issue: unconstrained training aligns shortcut groups with their prevalence, causing over‑confidence in one group and under‑confidence in another. They introduce two encoder‑agnostic methods—a regularizer and a post‑hoc recalibration—to equalize prevalence across shortcut groups, achieving significant performance gains on chest‑drain‑pneumothorax benchmarks with both fine‑tuned CNNs and frozen foundation‑model backbones.
By Mohamed Amine Kina, Eike Petersen
The paper investigates whether recent attention‑mechanism improvements—specifically gated attention, Kimi K3, Kimi Delta Attention, and Attention Residuals—effectively eliminate the attention‑sink problem when scaling language models to a one‑million‑token context window. Using a new diagnostic suite called SinkProbe, the authors evaluate sink mass, massive activation, position‑resolved recall, and the recency gap across four small models that vary only in token mixing and depth. Their findings show that the training objective, rather than the architecture, drives the emergence of attention sinks; gating did not replicate its previously reported benefits at the larger scale, and sink mass, activations, and positional bias behaved independently.
By Sara Rizwan, Samaanah Abdus Salam
FedGenSC introduces a federated generative semantic communication system that uses a global generator with local discriminators, a semantic prototype bank, and SNR‑conditioned generation to address instability, semantic drift, and channel‑agnostic issues in GAN‑based federated learning. Experiments on the Europarl dataset over Rayleigh fading channels show that FedGenSC outperforms the FedDeepSC baseline under non‑IID data, achieving up to a 58.2% relative improvement in BLEU‑1 at 18 dB. Ablation studies confirm that each component independently contributes to the overall performance gains.
By Rita Abou Fares, Razan Al Kakoun, Maher Nouiehed, Hadi Sarieddeen
The paper investigates the privacy risks inherent in auditing machine unlearning (MU) when the audit relies only on querying the model for behavioral signals. It shows that such generic audit schemes inevitably leak information about the retained data set, providing a geometric transfer theorem that bounds the distinguishability of retained set membership based on audit accuracy. The study also analyzes how the unlearned set, target sample, and query protocol influence the privacy‑audit transfer coefficient, with empirical evidence from both convex and non‑convex models supporting the theoretical findings.
By Liou Tang, James Joshi, Ashish Kundu
arXiv:2609.09662v1 Announce Type: cross
Abstract: Deploying Large Language Models (LLMs) directly on mobile platforms at the edge is gaining traction due to a myriad of benefits, such as increased pr...
By Weisi Yang, Stephen Xia
arXiv:2609.06059v1 Announce Type: new
Abstract: As large language models evolve from question-answering systems into general-purpose agents, evaluation must move beyond static answer correctness to a...
By Yu Liu, Zhilin Liu, Zhiwei Yang, Shaojie Zhang, Zheyuan Deng, Tingwei Huang, Zhenbo Luo, Lei Jiang, Yanbing Liu, Pei Fu
arXiv:2609.09890v1 Announce Type: new
Abstract: Multi-view diffusion models have shown strong performance in scenes with strong geometric priors and sparse semantics, such as indoor rooms or simple o...
By Qi Zhang, Yanyifan Wang, Weiyuan Zhang, Hui Huang
arXiv:2609.10261v1 Announce Type: new
Abstract: Multi-modal medical image segmentation leverages complementary diagnostic information, yet fusion can underperform single-modality baselines when spati...
By Yuchen Pei, Xiaoyu Hu, Yixiong Zou, Dingwen Hu, Hui Chu, Yutao Ma, Shijun Qiu, Gang Li
arXiv:2609.10317v1 Announce Type: new
Abstract: Streaming talking-head generation produces each frame as its driving audio arrives, yet fidelity and efficiency have so far pulled in opposite directio...
By Yanru An, Ruiyan Wang, Wenwu Wei, Rui Bu, Qi Wang, Hongwei Hu, Zhengxue Cheng, Rong Xie, Li Song, Wenjun Zhang
arXiv:2607.05205v2 Announce Type: replace
Abstract: Fast and reliable motion detection is essential for machine vision and autonomous systems operating in dynamic environments. This work integrates e...
By Qinbing Fu, Jingyu Huang, Yan Xie, Jigen Peng, Yuchao Tang
arXiv:2609.09945v1 Announce Type: cross
Abstract: Machine learning-based credit scoring is increasingly central to Peer-to-Peer (P2P) lending, yet its resilience to adversarial manipulation, where ap...
By Gijs A. F. Niewzwaag, Marijn G. S. Veth, Manuele Massei, Marcos R. Machado
ComplicitSplat is a novel black‑box attack that leverages 3D Gaussian Splatting (3DGS) shading to create viewpoint‑specific camouflage, embedding adversarial content into scene objects that is only visible from certain angles. The method does not require access to model architecture or weights and can successfully fool a range of popular object detectors—including single‑stage, multi‑stage, and transformer‑based models—on both real‑world physical objects and synthetic scenes. This demonstrates that downstream models using 3DGS are vulnerable to adversarial manipulation.
By Matthew Hull, Haoyang Yang, Pratham Mehta, Mansi Phute, Aeree Cho, Haorang Wang, Matthew Lau, Wenke Lee, Wilian Lunardi, Martin Andreoni, Duen Horng Chau
The paper introduces Vision-of-Thought (VoT), a framework that inserts a discrete visual-thinking layer between vision‑language models (VLMs) and diffusion transformers (DiTs). Instead of using VLMs solely as text encoders, they act as multimodal planners that generate VoT tokens—high‑level visual plans such as objects and layouts—before pixel rendering. A specialized VoT tokenizer is trained with a closed‑loop objective combining VLM alignment, feature reconstruction, and vector‑quantization losses, ensuring the tokens are semantically readable by the VLM while preserving necessary visual information. Experimental results show that VoT improves semantic alignment and offers a structured, interpretable interface for controllable generation.
By Jingxiang Sun, Chao Liao, Zhengxiong Luo, Chaorui Deng, Chen-lin Zhang, Junke Wang, Ceyuan Yang, Haoqi Fan, Weilin Huang
A*-Thought-V2 is a framework that models Chain-of-Thought reasoning as a geometric trajectory in a 3D PCA space, using explicit-implicit latent tokens to compress steps that deviate from the main question-to-solution direction. The method measures alignment angles to decide which steps remain text and which become latent, and introduces stepwise embedding forcing and label forcing to train the architecture. Experiments on Qwen models show up to 2.6% accuracy gains, halved response length, and significant reductions in computation and training time.
By Xiaoang Xu, Siyuan Liu, Shuo Wang, Junlan Feng, Fanyu Meng, Zhu Zhang, Jixun Wang, Xiaorong Wang, Zihan Zhou, Xin Li, Chaojun Xiao, Yiming Zhang, Huijia Wu, Liuyu Xiang, Peipei Li, Zhaofeng He
SAFIRE is a large-scale benchmark for fire and smoke understanding in multimodal large language models (MLLMs), featuring 83,000 captioned images across 20 scenarios and 193,000 multiple-choice VQA questions derived from a 9.7K-image subset. The benchmark evaluates 10 dimensions of performance, from basic perception to higher-order reasoning, and employs a GPT‑5.4-assisted verification pipeline to ensure annotation quality. Experiments on ten open-source MLLMs (8B–38B) reveal an average accuracy of 61.9%, highlighting significant gaps in safety-critical reasoning, while fine-tuning vision encoders on just 7% of SAFIRE data boosts fire-scene classification accuracy from 20.1% to 64.5%. All resources are publicly available at https://risys-lab.github.io/SAFIRE/.
By Pengfei Li, Naufal Suryanto, Sicheng Zhang, Mohammad Alsharid, Muzammal Naseer