PEPPER (PErcePtion-Guided PERturbation) is a backdoor defense for text-to-image diffusion models that rewrites input captions into semantically distant yet visually similar versions, adding unobtrusive elements to disrupt trigger embeddings. By escaping the attacked neighborhood in the text embedding space, PEPPER reduces attack success, especially against text encoder-based attacks, while preserving generation quality. The method requires no training or model weight access and can be combined with existing defenses to achieve stronger, generalizable robustness.
By Oscar Chew, Po-Yi Lu, Jayden Lin, Kuan-Hao Huang, Hsuan-Tien Lin
The paper introduces a new framework for predicting grapevine cold hardiness that learns transferable latent representations capturing region-specific variation. By inferring embeddings from cultivar descriptions, growing region text, and limited historical data, the method supports zero‑shot and few‑shot transfer to unseen regions. Experiments across six North American regions show that this approach outperforms existing models, delivering more accurate predictions and better performance in data‑scarce areas.
By William Solow, Paola Pesantez-Cabrera, Markus Keller, Lav Khot, Sandhya Saisubramanian, Alan Fern
The paper introduces the Frequency Selective Neural Network (FSNN), a new foundation architecture for time‑series learning that embeds advanced signal‑processing mathematics into its neural topology. By using a fully differentiable Wiener‑like filter bank optimized with complex‑domain backpropagation, FSNN autonomously discovers and isolates the precise physical modes of a given task, thereby avoiding the spectral entanglement that plagues CNNs, RNNs, and Transformers. Extensive evaluations show that FSNN achieves state‑of‑the‑art predictive performance, attaining 77.0 % average accuracy on the 10 multivariate UEA datasets and leading all major metrics on the imbalanced PTB‑XL ECG benchmark, while converging directly on physically meaningful frequency bands such as the cardiac QRS complex.
By Hui Huang, Ye Sun, Shiyan Hu
arXiv:2606.13061v3 Announce Type: replace
Abstract: Reasoning-driven universal multimodal embedding has advanced rapidly by introducing Chain-of-Thought (CoT) reasoning into the embedding pipeline. D...
By Peixi Wu, Biao Yang, Feipeng Ma, Bosong Chai, Bo Lin, Wei Yuan, Fan Yang, Tingting Gao, Hebei Li, Xiaoyan Sun
arXiv:2608.28840v1 Announce Type: new
Abstract: Independently trained neural networks tend to encode the same data with similar latent geometries. These latent geometries are not directly compatible,...
By Cameron Ryan, Vivek Sivaraman Narayanaswamy, Kowshik Thopalli, Shusen Liu
arXiv:2603.12824v3 Announce Type: replace-cross
Abstract: Vision-Language Model (VLM) based retrievers have advanced visual document retrieval (VDR) to impressive quality. They require the same multi...
By Zhuchenyang Liu, Yao Zhang, Yu Xiao
arXiv:2608.29748v1 Announce Type: new
Abstract: Speculative decoding accelerates autoregressive language model inference by having a lightweight draft model propose multiple candidate tokens, which a...
By Luxi Lin, Zhanpeng Zeng, Shuang Peng, Songwei Liu, Rongrong Ji
The paper introduces “Persuasio”, a multi‑agent dialogue platform that uses a formal argumentation theory to adjudicate winners in free‑text debates. Using this system, the authors generated 192 debates on a UK political topic involving humans and large language models (LLMs), and evaluated 22 interlocutors through automated adjudication and 9,702 crowdsourced pairwise judgments across 1,386 annotation instances. The results show a consistent decoupling between subjective persuasiveness—where LLMs dominate—and formal argumentative strength—where humans remain competitive, with multi‑agent and retrieval‑augmented variants widening this gap.
By Jordan Robinson, Angus R. Williams, Katie Atkinson, Anthony G. Cohn
arXiv:2608.30320v1 Announce Type: new
Abstract: We describe the architecture and ablations of Qwen3.8-Flash-Next, a sparse mixture-of-experts model with 125B parameters, 6B activated per token, and a...
By Zihan Qiu, Zekun Wang, Xiao Li, Yanpeng Li, Yang Xu, Yixuan Wang, Huaqing Zhang, Rui Men, Bochao Mao, Chengruidong Zhang, Fan Zhou, Hao Luo, Haofeng Huang, Haoran Lian, Haoyan Huang, Hongqing Chen, Jianwei Zhang, Jing Xu, Junjie Wang, Langshi Chen, Liangyu Wang, Linlang Jiang, Man Yuan, Minmin Sun, Peng Jin, Siqi Zhang, Siyu Wang, Xingzhang Ren, Yakai Wang, Yi Zhang, Yiming Dong, Yizhong Cao, Yubo Ma, Yunfei Mao, Bo Zheng, Dayiheng Liu
arXiv:2601.12075v2 Announce Type: replace
Abstract: Language models used in retrieval-augmented settings must arbitrate between parametric knowledge stored in their weights and contextual information...
By Mehrdad Farahani, Franziska Penzkofer, Richard Johansson
arXiv:2608.30996v1 Announce Type: new
Abstract: Retrieval-augmented generation systems can precompute and store key-value caches of retrieved documents to avoid re-encoding context at every query. Qu...
By Atta Ul Asad, Ahsan Bilal, Muhammad Ali, Muhammad Haseeb, Dean F. Hougen
arXiv:2608.28905v1 Announce Type: new
Abstract: Generative recommenders increasingly emit semantic IDs (SIDs): each item is a short sequence of hierarchical discrete codes from a residual quantizer,...
By Artem Betlei
arXiv:2608.29001v1 Announce Type: new
Abstract: In a data-driven world, efficiently organizing and mapping relationships between objects is crucial. Graphs are powerful tools for modeling these conne...
By Thiago C\'esar Castilho Almeida, Gustavo Rosseto Let\'icio, Lucas Pascotti Valem, Andr\'e Freitas, Daniel Carlos Guimar\~aes Pedronette
arXiv:2604.07146v3 Announce Type: replace
Abstract: Knowledge-based visual question answering (KB-VQA) requires vision-language models to understand images and use external knowledge, especially for...
By Zhuohong Chen, Zhenxian Wu, Yunyao Yu, Hangrui Xu, Zirui Liao, Zhifang Liu, Xiangwen Deng, Pen Jiao, Haoqian Wang
arXiv:2608.29239v1 Announce Type: new
Abstract: Low-resource ASR remains difficult because scarce transcripts provide limited supervised evidence for target-side generation. To address this gap, we p...
By Kuan-Tang Huang, Cheng-Yeh Yang, Chien-Chun Wang, Hung-Shin Lee, Hsin-Min Wang, Berlin Chen
arXiv:2608.29715v1 Announce Type: cross
Abstract: Transformers rely on position embedding mechanisms in long context modeling in most cases. Rotary Position Embedding (RoPE) embeds positional informa...
By Yixing Li, Ruobing Xie, Yudong Zhang, Yushi Bai, Samm Sun, Yu Cheng
The paper introduces a unified model that jointly performs Khmer text recognition and word segmentation, eliminating the need for a separate segmentation step. Using a connectionist-temporal-classification decoder, the model can be instructed to output Khmer text with or without word boundaries. Experiments across document, scene, and handwritten image datasets demonstrate that the model accurately recognizes characters and locates word boundaries, reducing error and latency compared to traditional sequential pipelines.
By Marry Kong, Rina Buoy, Sovisal Chenda, Nguonly Taing, Masakazu Iwamura, Koichi Kise
The paper introduces a configurable semantic chunking framework for biomedical information extraction in retrieval‑augmented generation systems. It replaces the fixed‑size chunking stage of BioMedRAG with entity‑preserving windows, trigger‑centered chunking, proposition‑first extraction, tiered trigger prioritization, and hierarchical relation resolution, while keeping the rest of the pipeline unchanged. Experiments on relation extraction benchmarks (GM‑CIHT, DDI, ChemProt) and adverse event classification (ADE) show that the hybrid configuration boosts performance on datasets with explicit relation cues, achieving 82.6% F1 on GM‑CIHT compared to 74.2% with the baseline.
By Riya Ahuja (Institute of Data Science in Biomedicine, TU Braunschweig, Braunschweig, Germany, Braunschweig Integrated Centre of Systems Biology, TU Braunschweig, Braunschweig, Germany), Tim Kacprowski (Institute of Data Science in Biomedicine, TU Braunschweig, Braunschweig, Germany, Braunschweig Integrated Centre of Systems Biology, TU Braunschweig, Braunschweig, Germany), Roya Shiasi Sardoabi (Institute of Data Science in Biomedicine, TU Braunschweig, Braunschweig, Germany, Braunschweig Integrated Centre of Systems Biology, TU Braunschweig, Braunschweig, Germany)
The paper investigates why token prediction, a common pre‑training objective for language models, yields useful representations. It introduces a statistical framework linking token prediction accuracy to the geometry of token embeddings, showing that accurate predictions organize embeddings according to Hellinger distances between context distributions. The authors also propose a self‑consistency principle that refines contextual representations through repeated application of a shared block, and provide downstream guarantees for token generation, community recovery, and linear classification.
By Shulei Wang
arXiv:2608.29953v1 Announce Type: new
Abstract: Flat retrieval-augmented generation treats a corpus as a bag of chunks, discarding document hierarchy and cross document structure. We introduce Search...
By Guransh Singh, Vishwajeet Kumar, Arkadeep Acharya, Adnan Qidwai, Jaydeep Sen, Sachindra Joshi