Physics-Transfer Learning (PT) is a physics-constrained learning theory that enables zero-shot generalization across domains of increasing complexity. PT builds chains of anchor domains (CADs) linking low-cost source domains to complex targets and introduces computable feasibility and reliability indicators to assess transfer before target labels are available. The authors derive an upper bound on target risk under physical-consistency and stable-representation assumptions, and demonstrate PT’s effectiveness on quantum excited-state energy prediction and classical dynamical-transition prediction, achieving accurate high-fidelity predictions across substantial distribution shifts.
By Yingjie Zhao, Zhiping Xu
The paper introduces ArchWorld, a benchmark of 3D architectural landmarks with extensive geographic coverage and detailed metadata, to evaluate and analyze the performance of 3D foundation models (3DFMs). Through a comprehensive error analysis, the authors uncover geographic disparities in model accuracy that stem from variations in data representation, scene size, and region-specific architectural features. They propose an intervention scheme that enhances 3DFM performance while simultaneously reducing these geographic disparities.
By Nicky Kriplani, Yiwen Zhang, Angelina Wang, Hadar Averbuch-Elor
The paper introduces Shift‑Reliance Alignment (SRA), a method for ranking vision‑language models (VLMs) based on how much they rely on specific cues such as scene context or text. By editing source images to mislead these cues and measuring model responses, SRA predicts which VLMs will be misled by physical sticker attacks and which will hallucinate on unedited photographs. In pre‑registered tests, SRA selects models that reduce error on challenging datasets like Objects365 and OpenImages, outperforming selection by source error and model confidence on the CueDial benchmark.
By Faisal Mohammad, Mohamed Younis
FederatedBytes is a small language model that processes text and multimodal data at the byte level, eliminating the need for tokenization. It uses a linear attention mechanism to keep memory and computational demands low, making it suitable for federated learning scenarios. Experiments on multilingual sentiment analysis tasks show that FederatedBytes outperforms models like ByT5 and LSTM, achieving higher accuracy while reducing memory and communication overheads.
By Ali Sarabadani, Massimo Piccardi, Mehran Abolhasan, Farzad Tofigh, Ian Zhou, Justin Lipman
The paper proposes a depth‑aware FrankenMerging technique that blends lower layers of an aligned large language model (LLM) with higher layers of its base counterpart to balance diversity and quality without extra training. Experiments demonstrate that adjusting the layer split controls both lexical and semantic diversity, often surpassing the aligned model in semantic diversity and outperforming temperature sampling in most settings. The method is validated by an ablation that reverses the merge order, supporting the hypothesis about layer‑specific diversity contributions.
By Yasuaki Sumita, Tomoharu Iwata, Toshiyuki Tanaka
The study investigates which layers of the self‑supervised speech model WavLM carry the most spoof‑relevant information for spoof detection. Using the ASVspoof 5 database, the authors perform per‑layer linear probing, canonical correlation analysis (CCA) with acoustic features, and a grid search over layer subsets, finding that five specific layers outperform pooling all thirteen. CCA shows mid‑layers align with most acoustic features, while voice‑quality metrics correlate more with later layers, and visualizations suggest deeper layers encode general synthesis artifacts rather than attack‑specific signatures.
By Mohamed Mallat, Michele Panariello, Massimiliano Todisco, Nicholas Evans, Anthony Larcher
The paper introduces an LLM‑enhanced framework for personalized stress detection that combines heart rate variability (HRV) features with personal attributes via semantic interpretation. It reports a 29.9% relative MCC improvement over the best baseline on the TILES‑2018 dataset and generates human‑readable descriptions of individual physiological responses. The approach demonstrates that large language models can provide both higher accuracy and interpretability in stress detection.
By Wei-Hsiang Ma, Ya-Tse Wu, Hsing-Hang Chou, Chi-Chun Lee
The paper analyzes primal‑dual algorithms for constrained Markov decision processes under a two‑timescale framework, focusing on how non‑stationarity enters via additive and parametric channels. In the additive case, oscillating constraint levels cause the dual variable to saturate at a floor determined by the moving optimum, with amplification proportional to the timescale ratio. In the parametric case, oscillating constraint sensitivity leads to a parametric resonance at twice the natural frequency, driven by mismatches between primal and dual estimates; this resonance is mitigated by aligning the updates or using a PID term.
By Maxime Baelde
The paper introduces task‑aware kernel Stein certificates for Rényi differential privacy, leveraging kernel Stein discrepancies based on escort distributions to provide privacy guarantees. By tailoring the kernel to focus on relevant privacy directions, the authors derive exact identities for Gaussian and Laplace mechanisms and bounds for exponential mechanisms, including task‑aware guarantees. A real‑world data example demonstrates how the protected comparison set and the released subspace jointly influence the resulting certificates.
By Leoni Carla Wirth, Tara Trautwein, Gesine Reinert
TRIP-Shapley is a propagation‑aware contribution attribution framework for peer‑to‑peer federated learning that avoids explicit multi‑round path tracking by computing round‑wise contribution scores and propagating only scalar values. This approach reduces communication costs by up to 45.8× compared to the DICE method while still attributing influence across the entire training process. The framework also includes lightweight defenses against misreporting attacks, which effectively lower contribution inflation in evaluated scenarios.
By Honoka Anada, Tatsuya Kaneko, Yafei Ou, Shinya Takamaeda-Yamazaki
The paper presents a pipeline for training, compiling, and deploying Bayesian neural networks (BNNs) that use a single probabilistic forward pass (PFP) instead of multiple stochastic samples. By assuming Gaussian-distributed weights and activations, PFP enables fully analytic uncertainty propagation, allowing a single deterministic forward pass to replace sampling. Using the TVM compiler and a library of Gaussian-propagating operators, the authors achieve up to 4200× speedups on small mini-batches while maintaining accuracy, uncertainty estimation, and out-of-domain detection comparable to traditional stochastic variational inference (SVI).
By Bernhard Klein, Falk Selker, Hendrik Borras, Sophie Steger, Franz Pernkopf, Holger Fr\"oning
The paper introduces a new family of Poisson non‑negative matrix factorization (NMF) methods that use a shifted‑log link function, allowing the model to transition from additive to more multiplicative part combinations via a single tuning parameter. An algorithm for maximum‑likelihood fitting and a computationally efficient approximation for large, sparse datasets are presented. Experiments on various real datasets demonstrate that the choice of link function can significantly affect results and may enhance interpretability compared to standard additive Poisson NMF.
By Eric Weine, Peter Carbonetto, Rafael A. Irizarry, Matthew Stephens
arXiv:2506. 19695v3 Announce Type: replace-cross Abstract: This paper studies the $\ell^p$-Lipschitz constants of ReLU neural networks $\Phi: \mathbb{R}^d \to \mathbb{R}$ with random parameters for $p \in [1,\infty]$.
By Sjoerd Dirksen, Patrick Finke, Paul Geuchen, Dominik St\"oger, Felix Voigtlaender
The paper investigates the tradeoff between fairness and accuracy when a single predictive model serves multiple demographic groups. It introduces the fairness-accuracy (FA) Pareto frontier, which identifies models that cannot be improved in either dimension without compromising the other. Focusing on finite-sample settings, the authors derive worst‑case‑optimal estimators, analyze how limited data asymmetrically affect each group’s welfare, and provide uniform bounds along the entire FA frontier.
By Alireza Fallah, Michael I. Jordan, Annie Ulichney
The paper examines how language models learn action nominal constructions (ANCs) across 96 synthetic target languages with varied grammatical features. Using GPT‑2‑small trained from scratch, the study evaluates models on minimal pairs covering 34 grammatical phenomena, revealing that models mirror typological patterns in basic clause configurations but diverge in complement systems and ANC strategies. The findings suggest that natural language distributions may arise from both general learning pressures and functional demands not fully captured by current model experiments.
By Iskar Deng, Shane Steinert-Threlkeld
The paper investigates whether uncertainty estimates can gauge hallucination risk in retrieval‑augmented generation (RAG). It finds that uncertainty is informative only when retrieved evidence constrains the answer: with good evidence, unsupported claims are less likely, whereas poor retrieval coverage lets the model rely on internal knowledge and assign high probability to unsupported claims. Experiments on HotpotQA and SQuAD across four instruction‑tuned models show that perplexity‑based uncertainty achieves AUROC scores of 0.81–0.91 for detecting targeted unsupported edits, while length‑only classifiers perform near chance. The study also demonstrates that adding a second passage can recover detection performance and that combining uncertainty with a semantic‑drift score further improves calibration.
By Rajeel Ahmad Ansari, Tautik Agrahari
The paper presents a Bengali hallucination detection system that uses response‑conditioned query rewriting, hybrid retrieval, and task‑specific language‑model judging. It employs six retrieval routes and a router that selects between a 27B textual judge and a 14B agent capable of reasoning or executing Python. The authors evaluate the system on the Olikbochon challenge, reporting a private macro‑F1 of 0.956 and detailing runtime, execution traces, and corpus overlap findings.
By Shamim Hasan, Samiya Ahmed Chowdhury
The paper introduces “KoLing”, a bias‑controlled adversarial benchmark of 50.9k Korean instances designed to suppress superficial cues while preserving semantic content, and evaluates both LLM‑generated and LLM‑modified text detection. It also proposes “KoLingProbe”, an interpretable detector grounded in theoretical linguistics that models Korean‑specific properties across multiple linguistic levels. Experiments demonstrate that KoLingProbe outperforms strong baselines on all detection tasks and remains robust to adversarial attacks, with analyses showing that integrating complementary linguistic features across multiple levels improves both performance and interpretability.
By SungHo Kim, SangKeun Lee
The paper investigates whether large models are necessary for speech brain‑computer interfaces (BCIs) and introduces BEST, a two‑stage framework that uses hierarchical intermediate representations and a compact decoder to map neural activity directly to text. BEST achieves low word error rates (6.19% and 1.62%) on the Brain‑to‑Text ’24 and ’25 benchmarks while using only 3.85% of the parameters of comparable methods. The study also highlights that decoder scaling yields non‑monotonic gains, cross‑modal alignment is more critical than model size, and overfitting harms sequence‑level performance.
By Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan, Ting Han
The paper introduces DarkLens, a lightweight method for detecting faithfulness hallucinations in large language models (LLMs). It estimates context reliance by projecting hidden states onto the dark space—directions weakly aligned with the unembedding matrix—and uses the resulting Dark Matter Ratio (DMR) as a direct score of context utilization. Combined with a parametric knowledge score, DarkLens operates within a single forward pass and outperforms many baselines on QA and summarization benchmarks across several open‑weight LLMs.
By Alexandra Kuleshova, Andrey Volodichev, Daria Kotova, Alexey Zaytsev