TH-GNN is a heterogeneous temporal graph neural network designed to detect shilling attacks generated by large language model (LLM) agents. It combines a two‑layer Heterogeneous Graph Transformer with per‑type and per‑relation attention, learnable sinusoidal temporal encodings, cross‑modal attention that fuses user embeddings with frozen RoBERTa representations of reviews and item descriptions, and a GRU that models log inter‑arrival times. Across five attack families and four benchmark datasets, TH‑GNN achieves a grand‑mean F1 score of 0.870, surpassing the best text‑only baseline on Agent4SR attacks by 10.9 percentage points and 11.5 percentage points at the lowest injection rate.
By Shivam Swarup, Divya Prakash Shrivastava, Rakesh Thakur
arXiv:2608.20868v1 Announce Type: cross
Abstract: Document processing pipelines traditionally cascade optical character recognition (OCR) engines with downstream models for structured information ext...
By A. Said Gurbuz (IBM Research Zurich, ETH Zurich), Ahmed Nassar (IBM Research Zurich), Christoph Auer (IBM Research Zurich), Maksym Lysak (IBM Research Zurich), Lucas Morin (IBM Research Zurich), Matteo Omenetti (IBM Research Zurich), Tim Strohmeyer (IBM Research Zurich), Panagiotis Vagenas (IBM Research Zurich), Nikolaos Livathinos (IBM Research Zurich), Michele Dolfi (IBM Research Zurich), Peter Staar (IBM Research Zurich)
The paper introduces Llama-Mobile, a framework that quantizes vision‑language models for efficient mobile deployment. It uses a quantization pipeline that generates training data from the model itself, eliminating the need for the original training setup, and employs a novel 2.7‑bit‑per‑parameter format optimized for Arm CPUs. Applying this method, the authors compress the Llama 3.2 11B Vision Instruct model to 3.7 GB with 8‑bit activations while maintaining strong performance on visual question answering tasks.
By Luka Ribar, Jeevan Bhoot, Douglas Orr
arXiv:2608.21099v1 Announce Type: cross
Abstract: Multi-modal object detection is essential for robust scene understanding in challenging conditions, including low-light and adverse environments. Rec...
By Jiekang Feng, Zhihe Fan, Yunqi Zhu, Xinjie Yao, Yueying Zhang, Yike Gao, Ranxin Li, Guanzuo Chen
The paper evaluates whether zero‑shot large language models (LLMs) can replace fine‑tuned natural language understanding (NLU) classifiers for intent detection in conversational systems. Experiments on ATIS and CLINC150 show that fine‑tuned RoBERTa outperforms Claude Haiku zero‑shot when ample in‑domain labels are available, but the LLM matches the supervised model on the broader 150‑intent CLINC150 schema. The authors identify three production‑relevant scenarios where LLMs excel—out‑of‑scope detection, robustness to ASR noise, and dynamic per‑deployment schemas—and present a decision framework to guide practitioners.
whyItMatters:"The study provides concrete evidence and a practical framework for deciding when to deploy LLMs versus fine‑tuned models in real‑world conversational applications."
By Carson Rodrigues, Oysturn Vas
arXiv:2608.20890v1 Announce Type: new
Abstract: Vision-Language-Action (VLA) models have emerged as a powerful paradigm for end-to-end autonomous driving by jointly integrating perception, reasoning,...
By Jingtao Sun, Xiaohai He, Yike Zhang, Dong Huang, Yaonan Wang, Ajmal Mian, Mike Zheng Shou
arXiv:2608.21343v1 Announce Type: cross
Abstract: Contextualization is essential for production automatic speech recognition (ASR) systems, where user-provided phrases must be recognized accurately u...
By Vladimir Bataev, Lilit Grigoryan, Andrei Andrusenko, Nikolay Karpov, Vitaly Lavrukhin, Boris Ginsburg
The paper investigates how visual presentation affects vision‑language models (VLMs) on the SPaRC spatial planning benchmark. By adding lightweight input‑side scaffolds that keep the visual modality but make spatial structure clearer, the authors achieve up to a 34.0‑percentage‑point accuracy boost across multiple VLMs, and an additional 4.6 points when combined with GRPO training. Analyses reveal that these improvements stem mainly from reduced grounding errors, while rule‑based reasoning remains difficult, highlighting visual presentation as a key determinant of whether VLM benchmarks test grounded perception, downstream reasoning, or both.
By Lars Benedikt Kaesberg, Tianyu Yang, Florian Valentin Wunderlich, Terry Ruas, Jan Philip Wahle, Daniel Kurzawe, Bela Gipp
SPARC (Single-Pass Adaptive Risk Calibration) is a Bayesian‑conformal uncertainty layer for human motion forecasting that adds an analytic epistemic scale to a deterministic MLP backbone’s Gaussian covariance. The scale, κ_t(x), inflates the covariance without altering its correlation structure, enabling 95% marginal prediction tubes with finite‑sample validity via split conformal calibration. Across nine dataset‑protocol blocks, SPARC outperforms baselines on NLL and a combined MPJPE+NLL metric while maintaining competitive point accuracy and efficient calibrated tubes.
By Sakif Hossain, Julian Teusch, J\"org P. M\"uller
arXiv:2608.21133v1 Announce Type: new
Abstract: Medical image-text data can expose protected health information (PHI) through both visible image content as well as accompanying text, creating a barri...
By Shiva Shrestha, Zongxing Xie, Chen Zhao, Liran Ma, Zhipeng Cai, Honghui Xu
The paper introduces CRATE, a two‑stage vision‑language model framework that evaluates mobile agents by reasoning about each step’s consequences and aggregating this evidence to assess task completion. It also presents CRATE‑S, an extension that evaluates operational safety. Experiments show CRATE and CRATE‑S outperform existing benchmarks, achieving high F1‑scores on AndroidWorld and MobileRisk datasets.
By Pengshuai Yang, Zijing Gao, Xue Yu, Benhui Zhuang, Bo Yuan, Junlan Feng
The paper introduces the Generative Embedding Benchmark (GEB), which evaluates how much content from an embedding can be recovered by a decoder that only has access to the frozen embedding and a question, without the original image or intermediate features. GEB uses a curated visual‑question‑answering dataset with 1,800 development and 900 test items covering natural images, scene text, and visual documents. Experiments on seven public embedding models show that visual‑only scores range from 28.25 to 33.21, while joint image‑question encoding boosts scores up to 65.56, revealing that generative readout uncovers information bottlenecks not captured by traditional separability‑based benchmarks.
By Yun Li, Biao Yang, Peixi Wu, Yunhao Zhou, Mingzhou Jiang, Wei Yuan, Fan Yang, Wenwu Ou
CogCanvas is a new benchmark for multi-subject reference-based image generation, featuring 1,952 curated reference images of 100 celebrities, 115 objects/fashion items, and 29 real-world backgrounds. It generates 1,361 compositional prompts with 2–5 people, using a pipeline that includes DINOv2 deduplication, aesthetic filtering, and automated graph derivation for interaction and positioning. The benchmark evaluates three tasks—reference-based multi-human-object generation, text-to-image compositional generation, and reference retrieval—under a six-axis protocol, and introduces BG‑Sim and Attr‑VQA metrics to assess background fidelity and attribute binding.
By Long-Bao Nguyen, Quang-Khai Le, Tam V. Nguyen, Minh-Triet Tran, Trung-Nghia Le
The study explores whether environmental data can enhance EEG-based affective-state classification by treating such data as literature-informed priors. Using a dual‑tower model that fuses EEG-Conformer representations with a graph‑based environmental encoder, the authors achieve 76.2% accuracy versus 67.4% for EEG alone on a dataset from Astana. Experiments with controls, dose‑response reversal, and domain‑shift show that the improvement is not solely due to environmental information, and replacing Astana’s environmental distribution with Singapore’s reduces accuracy to 72.8%.
By Utsav Poudel, Jagannath Aryal, Subramaniyaswamy Vairavasundaram
CertVLA is a certified defense for Vision‑Language‑Action (VLA) models that protects closed‑loop control against bounded patch and texture attacks. It calibrates a region of behaviorally consistent actions and uses deterministic covering masks to guarantee at least one attack‑free prediction, normalizing action disagreement by benign variation and accepting only consistent single‑mask anchors. The method provides finite‑sample clean coverage, extends the certificate to full rollouts, and proves that certified rollouts execute only action chunks consistent with clean predictions, ensuring task success regardless of patch content or physical transformation.
By Hui Lu, Zhijie Peng, Yuqi Lin, Zaijia Yang, Jiaming He, Shuhan Ye, Yi Yu, Hanwei Zhu, Bingquan Shen, Alex Kot, Xudong Jiang
The paper introduces Linear Discriminant Tree Ensembles (LDT, LDF, LDAB) for multimodal affect and behaviour classification, combining text, audio, and visual streams. It uses tokenization, concept clustering, and tree-based routing to balance accuracy and interpretability, and proposes a modified feature importance metric that reduces negative class bias. The ensembles outperform Multimodal Transformers and Interpretable Multimodal Routing in F1-mod and accuracy, and their feature importance aligns better with human annotations on IEMOCAP and CMU-MOSI datasets.
By Mojtaba Moattari
PhotoBench is a new benchmark built from authentic personal photo albums that moves beyond simple visual matching to focus on personalized, intent-driven retrieval. It incorporates a multi-source profiling framework that combines visual semantics, spatial‑temporal metadata, social identity, and temporal events to generate complex queries reflecting users’ life trajectories. Evaluation on PhotoBench reveals two key limitations: a modality gap where unified embedding models fail on non‑visual constraints, and a source fusion paradox where agentic systems struggle with tool orchestration.
By Tianyi Xu, Rong Shan, Junjie Wu, Jiadeng Huang, Teng Wang, Jiachen Zhu, Wenteng Chen, Minxin Tu, Quantao Dou, Zhaoxiang Wang, Changwang Zhang, Weinan Zhang, Jun Wang, Jianghao Lin
The article reviews how multimodal large language models (MLLMs) are expanding radiology AI beyond image‑specific tasks to multimodal reasoning, yet volumetric radiology poses a representational challenge because clinical interpretation needs full 3‑D spatial context and quantitative data. It surveys over 200 studies, categorizing advances in volumetric representation, multimodal understanding, and agentic orchestration, and introduces a Claim‑Design‑Validation framework to align technical, workflow, and clinical claims. The review emphasizes that native volumetric modeling and agentic capabilities must match spatial, quantitative, contextual, and workflow demands, and that clinical credibility hinges on faithful 3‑D representation, traceable behavior, proper validation, and defined human oversight.
By Zanting Ye, Shengyuan Liu, Xin Liu, Chenhui Wang, Zhisong Wang, Jiashuai Liu, Zipei Wang, Cheng Wang, Wentao Pan, Mengjie Fang, Di Dong, Mohammad Salmanpour, Arman Rahmim, Yu Gu, Yong Xia, Hongming Shan, Yixuan Yuan, Yefeng Zheng, Lijun Lu
arXiv:2608.21022v1 Announce Type: new
Abstract: Micro-actions are subtle, short, low-amplitude body movements, such as a fidgeting hand or a slight head tilt, that humans perform with little consciou...
By Fengshun Wang, Jin'ang Han, Zhigang Tu
arXiv:2608.20699v1 Announce Type: new
Abstract: Animating articulated 3D meshes via text requires satisfying strict kinematic constraints, modeling causal interactions between parts, and achieving in...
By Chunyu Zou, Peng Dai, Yi-Hua Huang, Ze Yuan, Jingwei Huang, Yeming Yao, Xiaojuan Qi