The paper evaluates how different chunking strategies—fixed-size, recursive paragraph-first, and content-defined—affect the cost of updating embeddings when Wikipedia documents are revised. Using 120 pairs of Chinese and English revisions, the authors find that fixed-size chunking requires the most recomputation (57‑68 % of the workload), while recursive and content-defined chunking need less (38‑56 % and 43‑54 % respectively). Despite these differences, evidence recall under a 2048‑token retrieval budget varies by less than two percentage points across methods and languages.
By Yuning Bian, Lele Xu, Yifan Yuan, Xinyu Yin, Chengyan Zhan
The paper proposes a method to improve open‑vocabulary semantic segmentation by refining the CLIP visual encoder. It introduces a selective attention mechanism that preserves regional semantics and adds visual prototypes—auxiliary memory‑based classifiers built from confidently predicted patches—to bridge the modality gap. An optimizable feature extractor with exponential moving average updates further enhances prediction quality, and experiments on five datasets confirm the approach’s effectiveness.
By Aijing Yu, Zhengbo Wang, Lijun Sheng, Jian Liang, Xiao-Yu Zhang
The paper introduces the Fabrication Severity Stack (FSS), a three-tiered framework that classifies fabricated citations by detection difficulty, ranging from non-existent references to subtle misattributions. It proposes a staged heuristic verification workflow that can be run through common large language model (LLM) interfaces, offering a zero-cost method for educators to check citations in dissertations and student reports. A pilot study on 100 citations, with 20% seeded fabrications, showed that mainstream LLMs could correctly flag 9–10 of 10 non-existent references and partially detect metadata corruption, indicating high recall without proprietary tools.
By Hazrat Ali, Mohamed Elawady
EHRAdapt is an adapter that maps electronic health record tuples directly into a frozen language model’s embedding space, using learned embeddings for modality, time gaps, and dedicated vectors for event codes. It represents each event vector as a sum of a frozen, ontology‑trained semantic prior and a low‑rank evidence residual, enabling robust learning even for rare clinical events. Experiments on four million patient records with three large language model backbones show that the full adapter outperforms all ablations in next‑event prediction and improves rare‑disease discrimination in classification tasks.
By Andre R Goncalves, Vincent Liu, Priyadip Ray
The paper investigates how phylogenetic leakage can inflate performance estimates for protein classifiers, particularly in venom ion‑channel target prediction. By treating cross‑species prediction as a structured distribution shift, the authors compare five structural inductive biases using a multimodal graph neural network that fuses ESM‑2 embeddings with a disulfide‑aware attention mechanism. Persistent‑homology representations, which capture backbone geometry through topological descriptors, outperform other biases under a leave‑one‑taxon‑out protocol, achieving 63.4% accuracy versus 54‑55% for other geometric priors and 30% for a sequence‑only baseline.
By Arezoo Bybordi, Favour Achimba, Mariam Gelashvili, Jessy Ramirez, Weigang Qiu, Mand\"e Holford, Anita Raja
The paper introduces Kernel-Anticipated Unlearning (KAU), a method for approximate unlearning of knowledge-graph embeddings that accounts for relearning during continued training. KAU uses the tangent kernel to model how forgotten facts’ scores change, expressing this response in closed form and guiding edits to minimize future leakage. Experiments show KAU reduces leakage by 46‑63% on FB15k-237 and 50‑82% on CoDEx-L compared to the strongest existing unlearner, while preserving higher utility.
By Parnian Fazel, Mohammad M Maheri, Hamed Haddadi
The paper studies random projections (RP) for spectral algorithms, providing a comprehensive analysis of their generalization performance under broad conditions. By exploiting the RKHS embedding property and operator similarity, it establishes optimal learning rates in Sobolev norms that match minimax lower bounds up to logarithmic factors. For common RP implementations—randomized sketches and Nyström sub-sampling—the required projection dimension scales with the effective dimension, reducing computational cost while preserving statistical efficiency.
By Caixing Wang, Yue Wang
The paper investigates how open‑book scoring evaluates document‑level LoRA adapters by training one adapter per Wikipedia article on Llama‑3.1‑8B‑Instruct. Results show that open‑book scores largely reflect a shared reading behaviour: a random adapter can recover most of a specialist’s advantage when the document is provided, but only a negligible fraction when the document is omitted in a cloze probe. While specialists improve cloze recall and their weight updates align with the documents they read, selecting adapters by weight‑space proximity or text embedding does not outperform random choice, and merged adapters recover most of the open‑book advantage but only a small portion of the cloze advantage. The study also documents four silent errors that yielded clean but incorrect results.
By Martino M. L. Pulici, Cuong Xuan Chu, Evgeny Kharlamov, Volker Tresp
The paper investigates whether EEG data can be treated as generic time series for classification tasks. It compares subject‑specific, subject‑agnostic, and subject‑conditional models, showing that off‑the‑shelf time‑series models trained per subject perform nearly as well as specialized EEG models, though not quite matching them. By incorporating subject embeddings into a joint subject‑conditional classifier, the authors achieve competitive performance on two of three datasets and even surpass all EEG methods on one dataset.
By Johannes Burchert, Thorben Werner, Vijaya Krishna Yalavarthi, Diego Coello de Portugal, Maximilian Stubbemann, Lars Schmidt-Thieme
The paper presents a method for compiling autoregressive large language model (LLM) control into a policy‑exact parallel inference scheme for Hierarchical Retrieval‑Augmented Generation (RAG). By constructing a shallow action‑prefix tree and using a single root prefill plus tree‑causal probes, the authors eliminate the need for serial decoding while preserving the exact control policy in exact arithmetic. Experiments on MetaX devices show that the compiled approach matches the serial policy on 84–89 % of questions and reduces control latency by 35–47 % compared to token‑wise serial decoding.
By Zehao Lin, Qianfeng Shen, Xixuan Hao, Shaobo Cui, Chunyu Li, Kai Chen, Bo Tang, Zhenjun Wang, Jin Yan, Zhiyu Li, Feiyu Xiong
The paper investigates how semantic meaning is arranged locally within large language models. By using two controlled probes—one cross‑lingual and one refinement probe—on four decoder‑only LLMs, it finds that after removing dominant variance directions, sentence pairs rank by semantic relatedness, semantic factors cluster around an invariant core with structured variation, refinement tightens neighborhoods, and sentence‑level versus token‑local readouts reveal complementary geometric patterns. These results suggest that LLM semantics form local anisotropic neighborhoods that contract when meaning is refined.
By Andr\'es Daniel Godoy Ortiz
The paper investigates whether language models can transfer hierarchical syntactic knowledge from a source language (L1) to a target language (L2) using a novel cross‑lingual poverty‑of‑stimulus (CPOS) experimental design. Results show that models can partially transfer L1 structural evidence to L2, producing a small but detectable hierarchical bias that varies across language pairs. The study further demonstrates that vocabulary overlap is crucial for successful transfer, indicating that cross‑lingual learning of hierarchical generalization is conditionally achievable.
By Shin Yamaguchi, Masaya Taniguchi, Tatsuki Kuribayashi, Taro Watanabe, Yoshinobu Kano
IndoNutriQA is a newly released Indonesian clinical nutrition question‑answer benchmark comprising 911 expert‑curated question‑answer pairs drawn from real clinical‑nutrition reports. Each pair is grounded in a clinically relevant context and enriched with structured metadata such as anthropometric assessment, medical diagnosis, and document provenance. The dataset supports generative, extractive, and retrieval‑augmented QA, and includes fine‑tuned baselines and zero‑shot evaluations of contemporary large language models.
By Tirana Noor Fatyanosa, Ummu Ditya Erliana, Mahendra Data, Muhammad Nurwegiono, Eleazar Tadeo Eman, Alham Fikri Aji
The study introduces an evidence-aware evaluation protocol for a retrieval‑augmented generation system focused on breast lesion management literature. Using an 18‑document corpus and 18 queries, the authors compared dense embeddings, BM25, and cross‑encoder reranking across various chunk sizes and overlaps, finding that a hybrid retriever with 384‑token chunks and 20% overlap achieved top‑three retrieval success for 17 of 18 queries and high recall and precision metrics. Expanding ground‑truth relevance annotations increased recognized evidence spans and accepted chunk targets, though retrieval rankings remained unchanged, underscoring the sensitivity of evaluation to annotation coverage.
By Miguel Mochizuki Silva, Samuel Ismael, Mateus C. Dias, Tha\'is Gaud\^encio, Yuri Malheiros, Barufaldi
arXiv:2610.30193v1 Announce Type: cross
Abstract: Retrieval-augmented generation (RAG) systems face security, privacy, and copyright risks including content leakage, source reconstruction, unauthoriz...
By Peiyang Liu, Xi Wang, Di Liang, Youru Li, Zhi Cai, Wei Ye
arXiv:2610.30485v1 Announce Type: cross
Abstract: Personal intelligence systems are taking on increasingly complex everyday tasks, yet repeated questions, unnecessary confirmations, and unwanted outr...
By Jialin Li, Zhenhao Chen, Hanjun Luo, Hanan Salam
arXiv:2610.30989v1 Announce Type: cross
Abstract: Personal agents need to draw on long-term interaction history to retrieve and apply external knowledge suited to a user's current request. Even an an...
By Wenjie Zhang, Zhichao Shi, Xuhui Jiang, Xiaojun Wu, Cehao Yang, Chengjin Xu, Xiaoning Song
arXiv:2610.30347v1 Announce Type: new
Abstract: We propose FiLM-CAM; a conditional-access image watermarking approach that enforces authorized encoding, decoding, and removal through keyed feature mo...
By Yash Kulthe, Vishal Asnani, Shruti Agarwal, Andrew Gilbert, John Collomosse
arXiv:2610.30724v1 Announce Type: new
Abstract: Video provides a natural source of supervision: what happens next serves as a learning signal for visual world modeling. Yet masked video pretraining f...
By Yifeng Yao, Jinghan Li, Zhihao Zhang, Yajiao Xiong, Yang Jin, Kun Xu, Huishuai Zhang, Dongyan Zhao
arXiv:2610.31284v1 Announce Type: new
Abstract: Long egocentric videos pose a central challenge for multimodal language models: the visual stream must be compressed into a memory representation that...
By Raffaele Cali, Giuseppe Lando, Rosario Forte, Antonino Furnari