# Research Papers — 2026-08 · bundle page 40 of 240 > Full-text markdown of papers #1951–2000 of 11995 published in 2026-08, newest first. AI-readable mirrors of the full 11995-paper 2026-08 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-08/page/40 [← Previous bundle](https://aipod.dev/papers/archive/2026-08/page/39.md) [Next bundle →](https://aipod.dev/papers/archive/2026-08/page/41.md) ### Algorithmic Principles For Multiclass Learning Are Hard To Come By: Limits of Regularization and Proper Learning Authors: Julian Asilis, Shaddin Dughmi, Vatsal Sharan, Alec Sun, Shang-Hua Teng, Chang Wang Published: 2026-08-27 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2608.26516v1 Two of the most fundamental questions in statistical learning theory are the following: which prediction problems are learnable, and how should they be learned? For the former, elegant answers often take the form of combinatorial dimensions. The latter question, however, has proved considerably more elusive: all known general-purpose multiclass learners rely on intricate orientations of exponentially large one-inclusion structures, and familiar algorithmic principles such as proper learning and regularization remain poorly understood. Motivated by prior work, we ask whether learning reduces to proper learning---possibly over a larger hypothesis class---and whether proper or improper multiclass learning can ultimately be captured by suitable regularizers. Our primary results answer both questions negatively, resolving three open problems from prior work. First, we exhibit a learnable multiclass problem that cannot be embedded in any properly learnable class, meaning learning cannot be reduced to proper learning by enlarging the hypothesis class. Second, we demonstrate that proper learning can require training error and characterize this phenomenon precisely: every properly learnable class admits a proper learner making $o(m)$ errors on samples of size $m$, but every prescribed sublinear scale $a_m=o(m)$ is necessary for some properly learnable problem. Third, regularization is not a general learner: we exhibit a properly learnable class that cannot be learned by any Structural Risk Minimization (SRM) learner, and a learnable class that cannot be learned by any local regularizer. We complement these impossibility results with a positive theory that gives two sufficient conditions for SRM learnability and characterizes SRM representability through integrability of revealed preferences. --- ### Sharp Minimax Regret for Infinite-Memory Logistic Prediction Authors: Vaneet Aggarwal Published: 2026-08-27 Categories: cs.IT, cs.LG arXiv: https://arxiv.org/abs/2608.26515v2 We determine the minimax cumulative log-loss regret of a finite-alphabet, exogenously driven source with genuinely infinite input memory: independent Rademacher inputs $(U_t)$ are observed sequentially and the next binary mark has logit $\sum_{j\ge1}θ_jU_{t+1-j}$, the unknown coefficients obeying a summable envelope $|θ_j|\le r_j$, $\sum_jr_j\le B$. At horizon $T$, lag $j$ can move the logit by at most $r_j$ and is exercised in only $n_{T,j}=(T-j+1)_+$ rounds, and the two limitations combine into the sum $Γ_T(r)=\sum_{j\le T}\log(1+n_{T,j}r_j^{2})$. One coordinate-localised Bayesian mixture achieves $R_T(r)\le CΓ_T(r)$ for \emph{every} summable envelope with $C$ universal. Our main result is a matching nonasymptotic converse for the canonical exponential and polynomial envelopes; its new ingredients are a modular finite-sample information bound for logistic experiments with an exogenous random design, and a conditioning estimate for the overlapping Toeplitz lag matrix obtained by exhibiting each off-diagonal Gram sum as a sum of independent Rademacher variables indexed by the edges of a forest, needing neither local asymptotic normality nor any spectral theorem for random Toeplitz matrices. So $Γ_T(r)$ is the minimax regret scale here, giving $Θ(α^{-1}\log^{2}T)$ for $r_j=Ae^{-αj}$ and $Θ(T^{1/(2s)})$ for $r_j=Aj^{-s}$, $s>1$ --- the latter without the extra $(\log T)^{1-1/(2s)}$ factor any window-truncation analysis pays. We also show memory decay cannot determine regret, and that a profile-scaled online Newton predictor attains $O_B(Γ_T(r))$ in polynomial time per round. --- ### Sycophancy Suppression Can Impair Rational Updating: Anti-Sycophancy Should Preserve the Ability to Update Authors: Huanhuan Ma, Henry Peng Zou, Chengze Li, Enze Ma, Yunyue Su, Philip S. Yu Published: 2026-08-27 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.26511v1 Large language models often exhibit sycophancy, revising their answers to align with users when users push back. Such answer flips, however, can arise from different causes. One possibility is that the model simply aligns with the user's feedback in order to satisfy them. Another is that the feedback genuinely contains useful evidence, prompting the model to update its answer in a rational way. We distinguish them as Unsupported-Yielding and Rational-Updating. Prior work focuses primarily on suppressing Unsupported-Yielding, while overlooking its effect on Rational-Updating. We address this gap with a two-turn evaluation framework that measures the two behaviors separately. Across representative training-time and inference-time interventions, we find that anti-sycophancy methods often encounter a trade-off in which reducing Unsupported-Yielding can sacrifice Rational-Updating, and vice versa, even when the two objectives are optimized jointly. Mechanistic analysis suggests that the two behaviors share an internal substrate: the MLP neurons and attention heads driving them overlap substantially, and their associated steering directions are positively aligned. We further conduct a preliminary orthogonalized steering exploration, which yields modest, backbone-dependent selectivity gains. Overall, our results suggest that anti-sycophancy should be treated not as a simple suppression problem, but as a selectivity problem, where effective interventions should preserve Rational-Updating while reducing Unsupported-Yielding. --- ### A Single Suffix to Break Them All: Basin-Aware Jailbreaks for Merged Model Families Authors: Yu Zhe, Yixin Tan, Junhao Wei, Wang Chen Published: 2026-08-27 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2608.26506v1 Model merging enables combining multiple fine-tuned models without additional training, but its safety implications remain poorly understood. Prior work primarily attributes merging risks to unsafe constituent models, implicitly assuming that merging individually aligned models preserves safety. In contrast, we show that model merging reveals a previously overlooked jailbreak risk rooted in the pretrained foundation model, even when all constituent models are individually safety-aligned. Motivated by this observation, we study a new threat setting where an attacker constructs jailbreak prompts that generalize across merged models sharing the same pretrained backbone, without access to the exact merging coefficients or constituent checkpoints. To exploit this phenomenon, we propose \textbf{Basin-Aware Jailbreak (BAJ)}, which formulates jailbreak generation as a min--max optimization over the merging space to produce transferable adversarial suffixes across merged model families. Experiments across diverse backbones and merging settings show that BAJ achieves consistently high transfer success rates and remains effective under existing defenses. --- ### NeuDonatello: Uncertainty-Aware Framework for Accurate Neural SDF Learning Authors: Alvin Jinsung Choi, Wanhee Kim, Taeyun Kim, Dasol Hong, Wooju Lee, Hyun Myung Published: 2026-08-27 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.26504v1 Neural surface reconstruction has emerged as a powerful paradigm for recovering high-quality 3D surfaces from multi-view images. However, recovering accurate geometry solely from RGB images remains challenging due to uncertainties arising from textureless regions, occlusions, and inherent scene ambiguities. Existing methods often overlook such uncertainties, leading to inaccurate estimates of the signed distance function (SDF). We introduce NeuDonatello, a novel framework that models and leverages SDF uncertainty to improve surface reconstruction. Central to our approach is to model spatially varying uncertainty using a Monte Carlo sampling strategy. Using this uncertainty, we develop an adaptive regularization that selectively strengthens geometric constraints where RGB supervision is unreliable, avoiding incorrect surface reconstruction. We further introduce an uncertainty-aware scale parameter for the SDF-to-density conversion. Conditioned on uncertainty, this design enables more accurate modeling of spatially varying densities. Extensive experiments demonstrate that NeuDonatello achieves state-of-the-art reconstruction accuracy, with robust performance across diverse scenes using only posed RGB images. --- ### Explicit Bounds on the Entropy of Piecewise Hölder Graphon Models Authors: Connor Loehde-Woolard, François G. Meyer Published: 2026-08-27 Categories: math.PR, cs.SI, stat.ML arXiv: https://arxiv.org/abs/2608.26501v1 We study the entropy of random graphs generated by piecewise Hölder continuous graphons. We first present a result on the rate of convergence of the normalized entropy as the size of the graph grows. The core ideas of the proof are described, with the detailed proof provided in the appendix. From this result, we then derive quantitative bounds on the entropy for the stochastic block model and random geometric graph model. These bounds provide explicit formulae rather than asymptotic statements which have been found previously. --- ### Systematic Literature Review of Machine Learning Models and Applications for Text Recognition Authors: Nuzhat Khan, Ab Al-Hadi Ab Rahman, Shahriyar Masud Rizvi, Ibrahim Yousef Alshareef, Muhammad Nadzir Marsono, Muhammad Paend Bakht, Mohd Shahrizal Rusli, Shahidatul Sadiah Published: 2026-08-27 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2608.26500v1 Optical Character Recognition (OCR) for text recognition using machine vision has significantly improved, particularly when handling heterogeneous textual data. Traditional OCR models struggle with script variations, writing styles, and degraded documents. Advancements in technology are leading to new AI models with improved architecture for handling multiple languages and complex data formats. Despite this progress, a comprehensive evaluation of OCR advancements remains limited. Based on the established preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines, this literature review presents an extensive assessment of OCR research to trace the evolution of AI models over the past decade. It explores the transition in AI models, application domains, data types, linguistic coverage, and challenges. Through a detailed analysis of 97 selected studies published during January 2015 - January 2025, key OCR models are identified, and their performance, strengths, and limitations are analyzed. The findings highlight how OCR technologies have evolved to address structured and unstructured text, scene text recognition, and multilingual processing. Unresolved challenges include limited resources for underrepresented languages, high variability in handwritten text, visual similarity among characters, and constraints in real-time OCR applications. To address these issues, several promising approaches are proposed. Key suggestions include self-supervised learning, multimodal AI, automated machine learning (AutoML), AI-assisted postprocessing, tiny machine learning (TinyML), and the creation of joint corpora for script matching. The future recommendations aim to enhance OCR accuracy and tackle the challenges identified for real-time industrial applications. This study will guide future research and establish a foundation for OCR field. --- ### RTNav: Towards Real-Time Zero-Shot Object Navigation Authors: Easop Lee, Lingyu Zhang, Boyuan Chen Published: 2026-08-27 Categories: cs.RO, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2608.26496v1 Navigation in unknown environments to find unforeseen objects has become increasingly feasible with capable vision and language foundation models. However, these models also introduce non-negligible inference latency, which becomes an important concern when agents must operate continuously in the real world. Most state-of-the-art methods are still developed in synchronous simulators, where the environment waits for the agent to act and inference time is effectively free. As a result, agents are often designed around the sequential execution of perception, reasoning, and action, with little regard for time constraints. Under real-time execution, where wall-clock time counts towards the task budget, the inefficiencies of these architectures become clear. We show that recent zero-shot object navigation methods suffer consistent performance degradation under such realistic timing conditions. Motivated by this observation, we propose RTNav, a simple but effective architecture that treats inference latency, asynchronous environment stepping, and bounded compute as explicit design considerations. Evaluated on real-time variants of HM3D-v1, HM3D-v2, and HM3D-OVON, RTNav improves the success rate by up to 11% and the Success weighted by Completion Time by up to 5.1 points over prior work. --- ### Video-FLAIR: Not Whether to Reason, But How Authors: Yogesh Kulkarni, Pooyan Fazli Published: 2026-08-27 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.26495v2 Multimodal queries can require different types of reasoning. Some can be answered via perceptual reasoning, extracting information directly from the visual signal, while others require compositional reasoning that combines observations or deliberative reasoning that evaluates competing hypotheses. However, many existing methods apply a uniform reasoning strategy across queries, leading to unnecessary computation on simple tasks and insufficient reasoning on complex ones. We introduce Video-FLAIR, a training framework that learns to select the appropriate reasoning mode for each query using reinforcement learning. During training, the model generates responses under all three modes for the same prompt, enabling direct comparison. A composite reward compares these responses to favor the most effective one based on correctness, grounding, and cost, while discouraging unsupported or misaligned deliberation. This yields a supervision signal for learning adaptive reasoning without per-query annotations. Video-FLAIR improves accuracy over the Qwen2.5-VL base model by +5.4 on MathVista, +4.8 on Video-Holmes, and +4.8 on Video-MMMU, while reducing average token usage to 95 compared to 417 for always-thinking baselines. --- ### A Unified Framework for Fair and Personalized Decentralized Learning under Communication Constraints Authors: Krishnendu S. Tharakan, Carlo Fischione Published: 2026-08-27 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.26493v2 Decentralized learning systems aim to collaboratively train models across multiple clients without relying on a central coordinator. While decentralization improves scalability, privacy, and robustness, it also exacerbates three fundamental challenges: statistical heterogeneity across clients, fairness in client-level performance, and stringent communication constraints. This raises a natural question: \emph{how fair can decentralized learning be under limited communication?} We address this question by presenting a unified framework for decentralized learning under communication constraints, bringing together graph-based personalization, agnostic fairness, and compressed event-triggered communication. Specifically, we propose a new algorithm DMFL-SQ, a decentralized multi-task learning algorithm that couples personalized model training over a communication graph with an agnostic mixture fairness objective, while reducing communication through sparsification, quantization, and event-triggered synchronization. We establish convergence guarantees for general non-convex objectives and show that DMFL-SQ achieves an $\mathcal{O}(T^{-1/2})$ rate in expected squared Moreau-envelope stationarity despite sparse, quantized, and event-triggered communication. We further derive PAC-Bayes generalization guarantees for the fairness-aware mixture objective. Experiments on CIFAR-10 and the real heterogeneous MUSMET EEG dataset demonstrate that DMFL-SQ substantially reduces communication while maintaining predictive performance and improving fairness across clients. Together, our theoretical and empirical results show that personalization, fairness, and communication efficiency can be jointly achieved in decentralized learning while preserving the dominant convergence rate. --- ### Bayesian methods and Markov chain Monte Carlo algorithms for curve reconstruction and point cloud data analysis Authors: Asir Intesar Tushar, Ioannis Sgouralis Published: 2026-08-27 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2608.26490v1 Point-cloud data routinely captured by modern imaging and sensor technologies provide detailed geometric descriptions of objects and environments, but their analysis is hindered by large data volumes, localization noise, and missing information. In addition, existing point-cloud reconstruction pipelines typically return a single best-fit structure without uncertainty quantification. We introduce a fully Bayesian framework for representing point-cloud data and reconstructing closed curves, in which observed points are modeled as noisy perturbations of latent locations constrained to lie on the underlying curve that is regularized by a non-parametric prior. Posterior inference in our framework is carried out using a series of Markov chain Monte Carlo samplers tailored to point-cloud characteristics. Numerical experiments, including synthetic examples and real-world LiDAR datasets, show accurate reconstructions and quantified uncertainty over the recovered curves. --- ### Learning Woody Clearing With Loss Alignment for Zero-Shot Regrowth and Woody Segmentation Authors: Kal Backman, Jared Wood, Adam Roff Published: 2026-08-27 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.26489v1 Detecting woody clearing is vital for managing biodiversity. Deep learning models can detect change in woody vegetation from bitemporal remote sensing imagery, however generated products may not meet end-user specifications due to unaligned loss definitions. Further limitations of deep learning models are the reliance on large datasets which can be difficult to attain for spatially rare and ambiguous events such as regrowth detection. In this work we train a model to detect woody change using bitemporal Sentinel-2 imagery consisting of 7 years' worth of annual imagery across the state of New South Wales, Australia. To align the objective of the model with end-user metrics, we introduce the loss scaling coefficient $α$ which transforms the objective to optimize for specific $F_β$ scores. Introducing $α$ was found to increase precision by 1.85x or recall by 1.12x. We propose input imagery augmentation and generation techniques that allow the woody change detection model to zero-shot transfer to regrowth and woody segmentation tasks. For woody segmentation, image generation techniques using activation maximization with low $α$ values for stability and image generation techniques derived from handcrafted features utilizing a mosaic of clearing patches and artificial trees for contextual grounding were found to outperform prior woody segmentation works of the study area, reducing the overall error by up to 18.2%. For zero-shot woody regrowth, creating pseudo-post and prior images resulted in the model achieving an F1 score of 0.845, creating a foundation for future regrowth detection work. --- ### Shared Actors Need Not Share Critics: Effects of Value Mismatch in Parallel Reinforcement Learning Authors: Zhenya Liu, Yang Meng, Zhuokai Zhao, Xuefeng Liu, Yuxin Chen Published: 2026-08-27 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.26481v1 When a single policy is trained in parallel across multiple environments of the same task, such as procedurally generated levels, randomized dynamics, or curricula, implementations commonly use one critic across all sampled environments. Yet different environments can assign different expected returns to the same input visible to the critic. A critic without environment information must then reconcile distinct value targets, systematically shifting the sampled advantages within individual environments. Using illustrative bandit models with multiple environments and a common optimal arm, we characterize how this value mismatch redistributes sampled policy updates, reinforcing unhelpful actions while attenuating or even reversing useful ones. The oracle processes using no baseline, the shared value, or the value specific to the sampled environment have the same mean logit update at a fixed policy and converge to the same optimal policy, yet their realized learning paths can differ sharply. The analysis motivates a minimal intervention: give only a logged environment index to the critic so that it can separate the value targets. Controlled CartPole and MuJoCo experiments expose the predicted shifted values, advantages, and performance gaps. In the more complex BipedalWalker and Procgen settings, the same intervention yields more stable learning and higher returns. Across all $16$ Procgen games, the multihead conditional critic improves aggregate normalized return on $600$ unseen levels per game by $40.8\%$. In conclusion, the theory identifies value mismatch as a direct mechanism through which critic sharing can degrade stochastic learning dynamics, not captured by scalar estimator variance alone, and the experiments show that conditioning on an index is broadly effective in parallel reinforcement learning. --- ### GVS5H: Zero-Shot Self-Orchestration with Ledger-Based Control for Improved LLM Coding Performance Authors: Victor Gao, Vida Khosrowshahi, Ali Khosrowshahi, Xihao Sun, Juhyun Lee, Simon, Lee Published: 2026-08-27 Categories: cs.MA, cs.AI, cs.CL, cs.SE arXiv: https://arxiv.org/abs/2608.26480v2 Frontier coding performance is typically attained with large, costly proprietary models. We introduce ledger-based zero-shot self-orchestration (GVS5H), a training-free method in which fresh instances of one model decompose problems and coordinate through a shared file system. Across eleven open and closed-weight models on the 100 latest hard LiveCodeBench problems, the method yields as much as 25.6 points improvement, boosting several cheaper models to frontier-level performance. Orchestrated Qwen3.8 Flash Next scores 93.0% against Fable 5's 90.4% at 9% of the cost, while the smaller Qwen3.8-27B reaches 92.4%. Gains are not universal: some models are unchanged or worse. Transcript analysis attributes the gain to decomposition and persistent context. Inference-time organization can reach or exceed frontier coding accuracy at a fraction of the cost on self-hostable weights. --- ### Zero-Shot Video Restoration and Enhancement with Text-to-Image Latent Diffusion Models and Multi-Modal References Authors: Cong Cao, Huanjing Yue, Xin Liu, Jingyu Yang Published: 2026-08-27 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.26476v1 Zero-shot image restoration methods with text-to-image latent diffusion models have achieved great success in universal image restoration tasks without training. However, applying them to video restoration will result in severe temporal flickering. In this paper, we propose a novel framework for zero-shot video restoration and enhancement which uses a text-to-image latent diffusion model and multi-modal references. Through the proposed dual prompt tuning inversion and sampling, the inference time can be reduced to nearly 1/3 of the original. The performance and temporal consistency can be also significantly stregthened. By using the proposed texture-aware video token merging, the temporal correlation between frames can be further utilized to improve the temporal consistency. We futher propose the referenced self-attention and referenced token merging to support image reference. Experimental results demonstrate the superiority of the proposed method in restoring and enhancing temporally consistent videos. --- ### DTD-VAE: Disentangled Temporal Dependencies VAE for Credit Risk Prediction Authors: Xiaobo Guo, Lu-an Dong, Yanbo Wang, Peng Zhang, Cai Zhi, Youru Li Published: 2026-08-27 Categories: q-fin.RM, stat.ML arXiv: https://arxiv.org/abs/2608.26473v2 Evaluating customer creditworthiness is crucial for retail banking operations, as it impacts marketing strategies, customer relationship management, and credit risk control. Traditional methods often struggle to capture complex temporal dependencies and extract pertinent information from customer data, crucial for accurate risk assessment. Specifically, they fail to differentiate between temporal patterns indicative of credit risk and those reflecting general customer behavior or preferences, leading to suboptimal risk predictions. In this study, we introduce the Disentangled Temporal Dependencies Variational Autoencoder (DTD-VAE), an advancement over conventional VAE, designed to disentangle temporal dependencies and distinguish credit risk-related features from past customer preferences. The feature inference module of the DTD-VAE incorporates an autoregressive temporal dependency learning mechanism that adeptly captures the temporal dependencies among latent variables, enriching the model's comprehension of the inherent data structure. Furthermore, the feature generative module utilizes an element-wise gating mechanism that assigns independent weights to each dimension of the expert models, enabling a finer-grained disentanglement of latent variables, particularly those relevant to credit risk prediction. Extensive experiments on six real-world datasets demonstrate that the proposed framework consistently outperforms existing methods, achieving performance gains of 3.2%-4.86% in ROC-AUC and 6.41%-9.71% in Accuracy Ratio. --- ### Seasonal and Quantum-inspired Models for Neutron Monitor Time Series Forecasting Authors: Krishna Bhatia, Shalini Devendrababu, Srinjoy Ganguly Published: 2026-08-26 Categories: cs.LG, quant-ph arXiv: https://arxiv.org/abs/2609.30281v1 We present a focused and reproducible study of multi-horizon forecasting on the Lomnicky Stit neutron monitor (LMKS) time series. Our evaluation suite covers simple seasonal baselines, modern deep sequence models, and functional and quantum-inspired architectures, including Seasonal Naive, Long Short-Term Memory (LSTM), Temporal Convolutional Network (TCN), N-BEATS, Kolmogorov-Arnold Networks (KAN), and two quantum-inspired variants, QiLSTM and QiKAN. We describe the dataset characteristics, diagnostic analysis, preprocessing pipeline, and training procedures, and report aggregate point-forecast performance using mean absolute error (MAE) and root mean squared error (RMSE) for all evaluated models. Our quick-run results indicate that the quantum-inspired KAN variant, QiKAN, achieves the lowest aggregate forecasting error among the evaluated configurations, while the simple Seasonal Naive baseline remains remarkably competitive. These results suggest that, for highly periodic scientific monitoring time series, models incorporating strong seasonal or low-dimensional functional priors can match or outperform substantially more complex sequence architectures. The findings motivate further investigation of parsimonious and decomposable function approximators for forecasting periodic scientific signals. --- ### DeepGOF-1: A Pretrained Convolutional Goodness-of-Fit Test for Logistic Regression with a Computable Consistency Certificate Authors: Ebrahim Khaled Ebrahim Published: 2026-08-26 Categories: stat.ME, stat.CO, stat.ML arXiv: https://arxiv.org/abs/2609.29575v1 Goodness-of-fit tests for logistic regression are least reliable where they are most needed: at small samples their levels drift from the nominal one, and combining them worsens the drift. We propose a test whose statistic is a convolutional network, trained once on simulated departures, that reads misfit as a picture: a grid of standardized residuals over covariate ranks. The analyst never trains. The network ships frozen, and the p-value is the rank of the observed score within the analyst's own bootstrap, so the level is a property of the calibration rather than of what the network learned. We prove exactness under pivotality, asymptotic exactness without it, and a consistency theorem whose key condition is computable from the frozen weights in one forward pass, giving a per-alternative certificate; we also measure the test's blind cone. On a pre-declared sixty-cell grid the deployed level stays in the nominal band in fifty-eight cells, while that grid's power criterion failed; on a published benchmark it is the most stable of thirteen levels across settings and sample sizes, and the test outpowers every partition test at every sample size while ranking seventh overall. A bank-failure application and a versioned release close the paper. --- ### ModularSQL: A Runtime Guardrail for the Multiplicity Blind Spot in Text-to-SQL Authors: Tianxin Zhou, Ruixi Lin Published: 2026-08-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.29573v1 Text-to-SQL systems are increasingly deployed on production databases, where queries that pass benchmark evaluation can still produce results that distort downstream workflows. Standard set-based execution accuracy (Set-EX) collapses duplicate rows and can therefore miss multiplicity errors, including missing DISTINCT, inflated aggregates, and Cartesian-style join explosions. We call this the Multiplicity Blind Spot (MBS) and introduce Multiset-EX, a multiplicity-preserving evaluation criterion that exposes such failures. Across released DeepEye-SQL artifacts from three backbones (Qwen2.5-Coder-32B, Qwen3-Coder-30B-A3B, and Gemma-3-27B) on executable BIRD-Dev N=1532, we find a consistent 5.81--6.79 pp gap between Set-EX and Multiset-EX. The gap is not specific to DeepEye-SQL: it persists on released DAIL-SQL+GPT-4 (5.22 pp) and BIRD GPT-3.5-turbo (3.39 pp) predictions. We further introduce ModularSQL, a lightweight post-selection runtime guardrail that probes executed results for multiplicity anomalies and applies deterministic patches or low-cost LLM rescue only to flagged queries. Integrated with DeepEye-SQL using Qwen3-Coder, ModularSQL preserves Set-EX at 72.06% while improving Multiset-EX from 65.86% to 67.75% (+1.89 pp). It flags 77 high-risk anomalies, while adding only $0.0076 in total LLM cost and 120 ms amortized latency per query. Cross-pipeline evaluation shows that the candidate-free detector and deterministic patches also transfer to independently released prediction sets. Overall, these results show that benchmark accuracy does not necessarily imply execution-safe SQL, and that lightweight, multiplicity-aware runtime guardrails can narrow this gap with modest computational overhead. --- ### Task-Resolved Fisher Spectroscopy for Quantum Reservoir Computing Authors: Yang Peng Published: 2026-08-26 Categories: quant-ph, cs.LG arXiv: https://arxiv.org/abs/2609.29570v1 Quantum reservoir computing (QRC) uses fixed quantum dynamics to encode a time series and trains only a classical readout, but a benchmark capacity alone does not reveal whether task information is lost in the reservoir, the measurement, feature compression, or finite sampling. We introduce \emph{task-resolved Fisher spectroscopy}, in which prediction targets define orthonormal score coordinates on the stationary distribution of input histories. Reweighting labeled histories along these scores generates an exactly affine family of reservoir states and measurement outcomes. In the same coordinates, we obtain a many-body Fisher-information hierarchy relating the state quantum Fisher information, the Fisher information of the complete measurement record, and moment matrices retained through many-body order $r$. The quadratic form of each moment matrix is exactly the stationary capacity of the optimal linear readout, while a finite-measurement extension predicts the approach from one-shot records to ideal expectation-value features. The measurement-level quantities require only stationary labeled records and measured outcome strings, not an input model or quantum-state tomography. In a five-spin open reservoir, interactions route fourth-order temporal information into higher-body correlations, so low-order compression can incur orders-of-magnitude sampling overhead even when a conventional parity target remains accessible. For correlated inputs and outputs, record-defined task scores predict held-out capacities, the required feature order, and measurement-budget dependence; optimizing the local measurement axis recovers otherwise hidden task information. The framework connects benchmark performance to experimentally actionable diagnoses of reservoir encoding, measurement choice, classical representation, and shot allocation. --- ### Classifier-Dependent Benefits of Pseudo-Labeling for Semi-Supervised Android Malware Attribution Authors: Md Rafid Islam, Zahid Hasan, Hafiz Abdur Rahman Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.29564v1 Detecting and classifying Android malware families remains challenging due to high feature dimensionality, class imbalance, and the high cost of expert-labeled data. Semi-supervised learning (SSL) offers a way to leverage unlabeled samples, but prior works rarely test whether SSL benefits generalize across classifier types or report statistical significance. We present a systematic evaluation of pseudo-labeling across six classifiers (LightGBM, XGBoost, Random Forest, Logistic Regression, MLP, and SVM) on the CICMalDroid 2020 dataset, using five-fold stratified cross-validation and paired t-tests across five labeled ratios (1-20%). We find that SSL benefit is strongly classifier-dependent: SVM shows the largest significant gain (+4.4% accuracy at 5% labels, p = 0.0028), LightGBM improves modestly (+0.8 to +1.3% at 2-5% labels), while Random Forest is significantly harmed at low label ratios (-3.1% at 1% labels). Per-class analysis reveals SSL disproportionately benefits the hardest-to-classify families, with Adware F1 improving by +13.8 percentage points versus only +0.8 for the already well-classified Benign class. We further show that approximately 800 labeled samples (10% of the dataset) yield near-optimal performance across all classifiers. These findings offer practical guidance on when and with which classifier pseudo-labeling is worthwhile for Android malware classification. --- ### Is Reasoning Always Useful? Rethinking Reasoning Utility in Universal Multimodal Embeddings Authors: Wenxiao Fan, Jingling Fu, Luohang Liu, Xinyuan Shan, Lichen Ma, Yu He, Junshi Huang, Yan Li, Kan Li Published: 2026-08-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.29560v1 Reasoning-enhanced universal multimodal embeddings (UME) improve heterogeneous retrieval, but plausible rationales do not necessarily produce discriminative rankings. We study this gap by comparing the discriminative (DISC) and reasoning-driven generative (GEN) branches of UME-R1, a state-of-the-art reasoning UME method. We decompose reasoning utility into positive-target gain, hard-negative gain, and their margin difference. Positive similarity increases for 56.6%, but 15.7% are false-helpful cases where reasoning moves hard negatives closer even more. Local-neighborhood and token-attribution diagnostics suggest why: reasoning often de-condenses retrieved neighborhoods, but utility requires separator-aligned movement, while influential CoT tokens frequently encode evidence shared by positives and hard negatives. Motivated by these diagnostics, we propose SURE (Score-structure Utility Router for Embeddings), which improves UME-R1-7B by 1.5 points and yields consistent gains on two additional embedding models on MMEB-V2, without retraining, label-based policy selection, or extra VLM forward passes. --- ### Benchmarking Arabic--Russian Machine Translation: A Comparison of Fine-tuned NMT and Few-shot LLMs under Rich Morphology and Low Lexical Overlap Authors: Mullosharaf K. Arabov Published: 2026-08-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.29559v1 Arabic-Russian machine translation (MT) remains under-explored due to the rich morphology of Arabic and low lexical overlap between the two languages. We benchmark seven fine-tuned neural machine translation (NMT) models against four few-shot large language models (LLMs) on a 20k/5k/5k split of a new 15.47M-pair corpus. Fine-tuned NLLB-1.3B achieves the highest BLEU (16.3) and COMET (0.738). Aya-Expanse 8B leads the few-shot LLMs (BLEU 1.7 on 500 sentences, chrF 25.7), but all LLM scores remain far below the fine-tuned NMT baselines. Error analysis identifies low lexical overlap as the dominant failure mode; among the worst translations, mT5-small produces 32% too-short outputs. Bootstrap tests confirm significant differences among most models. Our results demonstrate that fine-tuned NMT significantly outperforms few-shot LLMs for Arabic-Russian translation under low-resource conditions. --- ### Visual Representation and History Modeling for Navigation World Models Authors: Guangfu Guo, Xiaoqian Lu, Rui Liu, Yutong Chen, Kunpeng Liu, Long Cheng Published: 2026-08-26 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.29555v1 Navigation World Models (NWMs) predict action-conditioned visual futures for planning. Two practical challenges are central to their design: selecting a suitable visual representation and efficiently modeling observation history for repeated candidate queries. Standard Global-Softmax attention provides flexible interactions but repeatedly processes the same history, leading to increasing computation and memory costs for long contexts and multi-query planning. We study both problems within a unified conditional flow-transformer framework. We first compare five frozen visual representations under the same dynamics model and evaluation. To reduce redundant history computation, we design Cached-Linear, a hybrid architecture that combines local and shifted-window attention for target mixing with linear attention for reusable history access. We further develop Balanced Gated Delta Network (GDN), which augments this design with frame-wise recurrent memory for temporal history modeling. Experiments on RECON, SACSoN, and SCAND show that representation choice depends on the prediction objective: PAE-L performs best for reconstruction, RAE-B for direct prediction, and V-JEPA for long-horizon rollout. Under shared-history workloads, Cached-Linear substantially reduces computation and memory compared with Global-Softmax, while Balanced GDN improves selected direct-prediction endpoints with efficient context reuse. Overall, we systematically study visual representation and history modeling for NWMs and develop hybrid reusable-history architectures for efficient long-context and multi-query prediction. --- ### UNWIND: Any-Length Facial Video for Stress Detection without Temporal Windowing Authors: Stefanos Gkikas, Christian Arzate Cruz, Eric Nichols, Giorgos Giannakakis, Randy Gomez Published: 2026-08-26 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.29553v1 Automatic stress recognition from facial video provides a non-contact approach for affective monitoring. However, most existing video-based methods divide complete recordings into shorter temporal segments before performing classification. Such segmentation requires additional decisions concerning segment duration, overlap, and prediction aggregation, and may restrict the model from exploiting information distributed across the entire recording. We introduce UNWIND, a facial-video framework for stress detection that analyzes a complete recording as a single model input, eliminating the need for temporal windowing or external segmentation. UNWIND reorganizes the video by folding its temporal dimension into the channel dimension of a two-dimensional spatial representation, which is subsequently processed through a unified asymmetric-attention architecture. With a temporal stride of $τ=1$, the framework processes the entire $120$-second sequence, corresponding to $3{,}600$ frames sampled at $30$~fps, in a single input. We evaluate seven temporal-stride settings on a stress dataset comprising $58$ subjects, using a stratified subject-level protocol that covers configurations from dense frame retention to sparse temporal sampling. The highest test accuracy, $70.02\%$, is obtained at $τ=15$, while processing all frames at $τ=1$ achieves a comparable accuracy of $69.73\%$. Computational requirements range from $12.48$ to $348.78$ GFLOPs across the evaluated stride settings, illustrating the balance between temporal sampling density and computational efficiency. The findings show that effective facial-video stress recognition can be achieved without dividing recordings into temporal windows and that complete-recording inference can be performed within a single unified model. --- ### HiPACE: Hierarchical Phase-Boundary Analysis and Controlled Evaluation of Feature Absorption in Sparse Autoencoders Authors: Jinyuan Zhang, Peng He, Yin Yuan, He Hu, ShengShuo Jiao Published: 2026-08-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.29551v1 Sparse autoencoders (SAEs) decompose LLM activations into sparse dictionary atoms, so that each distinct concept gets its own feature. One recurring behavior complicates this premise: feature absorption, in which a parent concept and its children--fruit and {apple, banana, pear}, say--collapse into a shared family direction. Prior work documents absorption empirically; missing is a closed-form prediction of when the shared direction is the cost-optimal representation of an active semantic family. This paper closes that gap. For a hierarchical Bernoulli generator with $k$ active children and residual scale $α$, the $L_0$-penalized reconstruction objective admits a closed-form phase boundary $λ_c(k,α)=α^2 k/(k-1)$: above it, pure parent absorption is strictly cheaper than pure child coding. Building on this boundary, we introduce HiPACE, an evaluation protocol that tests the boundary's structural consequence in real SAE dictionaries--measuring parent--child decoder structure over WordNet families, freezing the discovery-selected statistic before testing on unseen families, and contrasting genuine families against randomized sibling nulls. The boundary proves sharp in its native regime, predicting the synthetic transition within $\pm15%$ on all 30 tested cells. In Pythia-160m SAEs, the parent--child decoder gap recovers the predicted ordering with partial correlations up to $-0.93$ that sustain on the locked holdout and exclude sibling nulls ($p=0.002$). Controlled activation composition connects the theory's active-child count to the recovered family directions, and residual-stream interventions show that signed family directions increase parent-category logits, reversing under sign flip and vanishing under random controls--establishing causal sufficiency at the family-subspace level. --- ### Spaceborne differential photogrammetry for control-free measurement of large-gradient deformation with structural immunity and a predictable accuracy envelope Authors: Yueqiang Zhang, Chang Ma, Shuixin Pan, Haibo Liu Published: 2026-08-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.29550v1 Optical satellite image correlation measures wide-area deformation in regimes where coherent interferometric synthetic aperture radar fails because displacement gradients are too large. However, standard pairwise workflows lack a pre-acquisition error budget and rely on extensive stable terrain. We formulate repeat-pass optical correlation as a differential estimation problem without surveyed ground control. Nominal georeferencing defines the coordinate frame, stable-area constraints and displacement priors resolve the datum, and surface displacement is estimated jointly with inter-epoch revisit-bias coefficients. The model yields a predictive accuracy envelope and calibrated per-point posterior uncertainty, bounds along-track uncertainty through a displacement prior, and represents pushbroom jitter using per-line revisit offsets. Simulations and Sentinel-2 and WorldView-2 experiments on the 2019 Ridgecrest earthquake, the 2023 Kahramanmaraş earthquake, and the Baltoro glacier validate the predicted noise floor, control-free accuracy margin, and leakage caused by view-angle and digital elevation model errors. The measured noise floor reaches approximately $0.05$ pixel at $10$,m ground sampling distance. With only five stable tiles, conventional destriping changes the estimated Baltoro trunk velocity from $106$ to $1251$myr$^{-1}$, whereas the prior-constrained estimate remains $87$myr$^{-1}$. Closure analysis attributes approximately $88\%$ of pair-error variance to individual scenes, consistent with $25{,}354$ ITS_LIVE glacier-velocity triplets. Three matching methods lead to the same conclusions. The framework therefore turns pairwise correlation into a robust measurement with a predictive error budget, reduced dependence on stable terrain, and conclusions independent of the matching method. --- ### StepCOPS: Closed-Testing Lower-Tail Certificates for Language-Model Policy Selection Authors: Ibne Farabi Shihab, Sanjeda Akter, Anuj Sharma Published: 2026-08-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.29549v1 Post-training pipelines must select one language-model policy from many checkpoints, prompts, and decoding rules. Mean evaluator scores can conceal rare failures, whereas simultaneous candidate-wise confidence bounds can be unnecessarily conservative. We introduce StepCOPS, which uses an independent proposal split to nominate one lower-tail floor per candidate, exact binomial tests on a fresh certification split, and Holm's step-down procedure to certify a set of floors. With probability at least $1-δ$, every certified floor, including the largest floor used for policy selection, is below its candidate's population lower $α$-quantile. This guarantee assumes i.i.d. evaluation units while allowing arbitrary within-unit dependence across candidates. Across 24 predeclared configurations and 11 benchmarks, StepCOPS obtains 96.4% selected-policy coverage over 500 paired trials, raises the certified floor by 1.5 points over both proposal-Bonferroni and exact COPS, remains 0.6 points below the large-reference jury oracle, and abstains in 2.4% of trials. Shadow-judge, benchmark-native, artifact, and leave-one-judge-out audits characterize the proxy boundary: the guarantee applies to the fixed jury score, not directly to human safety. --- ### Certified Predictive Value-of-Advice Gating for Cost-Aware Language-Model Guidance in Reinforcement Learning Authors: Ibne Farabi Shihab, Md Najmus Swaqeeb, Abu Sa-Adat Mohamed Moon-Im Al Ahsan Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.29548v1 Language-model advice can accelerate reinforcement learning, but calls are costly and returned actions may be stale or wrong. We formulate advice acquisition as a response-contingent metareasoning problem: before querying, the controller predicts possible parsed responses, evaluates the decision and declared continuation that would follow each response, and queries only when a lower confidence bound on predictive value exceeds the priced cost. Execution is governed separately by an action-specific certificate. Under explicit assumptions, certified advice is near-optimal, a wrapped learner inherits fallback regret only under intervention stability, and conservative allocation loses at most the declared query-value estimation error relative to a myopic oracle. On BabyAI, a proxy-calibrated controller with Qwen2.5-1.5B and 7B advisors improves GoToObj return over no querying by 0.029 +/- 0.016 and 0.030 +/- 0.015 across 20 seeds while reducing calls by more than 97% relative to always-query. GoToLocal is a null result. Exactly matched-call tests show an advantage over random placement only for the 1.5B advisor and no advantage over an equal-budget early schedule. Mondrian calibration improves decision-relevant empirical coverage from 0.47 to 0.85, still below the 0.90 target, while the formally covered radius is vacuous. The demonstrated benefit is therefore robust sparse advice volume on a useful task, not a proven per-state placement advantage. --- ### Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets Authors: Olga Manakina, Igor Bogdanov, Chung-Horng Lung Published: 2026-08-26 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.29509v1 Large language models (LLMs) are increasingly deployed as multi-turn agents that must sustain goals, use tools, and adapt to other agents over extended interactions. However, existing research lacks auditable, multi-turn, multi-factorial experiments that quantify LLM behavior under explicit constraints, with time-resolved statistics that reveal how behavior unfolds over long horizons. To address this gap, we develop a multi-agent micro-benchmark inspired by the Stanford marshmallow experiment: ReAct agents operate minute-by-minute with a "raise a question" tool under a per-step budget, while we factorially manipulate social context (broadcast vs. isolated), personas (age, hedonic drive), and metacognitive policy (mandatory vs. optional tool use). We analyze outcomes with Kaplan-Meier (KM) survival curves and discrete-time hazard models over a long risk horizon across 19,200 agent trajectories in 64 cells. Behavior shows a sharp early "eat" impulse, and only 75.9% of agents persist to the end. In a discrete-time hazard model, isolation reduces per-minute risk relative to broadcast, whereas a must-use self-questioning policy increases risk. On average, agents ask $\approx 7.12$ questions and hit the per-step budget in $\approx 6\%$ of minutes. Questioning declines faster under broadcast than isolation. Ablation experiments demonstrated that removing hedonic drive and/or persona age increases survival and completion, narrows the broadcast/isolated gap, but leaves the must vs. may ordering intact. The combined ablation (no hedonic + no persona age) yields the highest completion (approaching $1.0$). These results establish delay-of-gratification as a compact, multi-turn interaction benchmark that captures social contagion and tool-use dynamics in LLM agents, providing a reproducible testbed and statistics for analyzing long-horizon, multi-agent behavior. --- ### Evaluation of Multi-Turn Consistency in LLM Agents: Survival Analysis and Failure-Rationale Taxonomy Authors: Igor Bogdanov, Olga Manakina, Chung-Horng Lung Published: 2026-08-26 Categories: cs.AI, cs.CL, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2609.29508v1 Large language model (LLM) agents may perform well on isolated tasks yet drift into inconsistency over extended interaction. We evaluate temporal consistency in a controlled 20-step multi-agent setting inspired by delayed-gratification studies. At each step, an agent chooses between continuing to delay a reward or claiming it immediately (terminating the episode). Across a full-factorial manipulation of social visibility (private vs public), persona stressors, and deliberation policy, we run 84,540 trajectories spanning 8 model families. Treating the first reward-claim as a time-to-event outcome, we estimate Kaplan-Meier survival curves and fit discrete-time hazard regression to quantify how experimental factors shift failure risk over time. Then, to analyze rationales and language patterns associated with failure, we build a seven-category taxonomy from 13,780 deliberation traces from agents who choose to terminate the episode, using an LLM-assisted labeling paired with human audit ($κ=0.83$). Rationale profiles change systematically with time and context: early failures are more impulse-driven, later failures more fatigue- and cost-benefit-framed, while public settings increase norm-oriented justifications. We also find a deliberation-inconsistency association: among failures, longer deliberation correlates with higher rates of intra-rationale contradiction (simultaneous pro-delay and pro-claim statements), challenging the assumption that more reasoning text implies greater consistency. Together, the survival and rationale analyses reveal distinct temporal reliability regimes and model-specific "failure fingerprints", offering an evaluation lens for diagnosing inconsistency in multi-turn agent behavior. --- ### Evaluating Explanation-Driven Vision-Language Reasoning via Generation Order Interventions Authors: Siting Liang, Luca Rippe, Omar Adjali, Daniel Sonntag Published: 2026-08-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.29496v1 Natural language explanation generation serves as a key mechanism for exposing and evaluating vision-language reasoning. Prior work on explanation-driven vision-language models predominantly follows a post-hoc (answer-first) paradigm, implicitly suggesting that supervised rationales can reflect underlying reasoning processes. In contrast, modern large vision-language models increasingly exhibit a rationale-first generation tendency, which more closely aligns with structured, stepwise reasoning. In this work, we systematically evaluate whether explanations are causally tied to model predictions within a single generation step under a controlled experimental setup, explicitly eliminating unnecessary chain-of-thought or other intermediate reasoning processes across knowledge-intensive QA, visual entailment, and compositional grounding benchmarks. We find that larger models emerge as a prerequisite for reliably supporting rationale-first reasoning at scale. However, answer-first generation is less prone to format-related errors in structured output. Overall, explanation ordering, model scale and pre-training knowledge, task-specific fine-tuning, and task structure jointly influence both prediction accuracy and reasoning faithfulness. --- ### BLADE: Distilled LLM Regularization for Calibrated Knowledge Graph Completion Authors: Ibne Farabi Shihab, Rabeya Bosri Tamanna, Abdo El Karaky, Sanjeda Akter, Anuj Sharma Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.29487v1 Knowledge graph completion models optimize ranking, although many downstream applications require calibrated probabilities. We present BLADE, a variational model that separates latent truth from graph recording and distills offline language-model judgments into a frozen teacher regularizer. The LLM is absent during inference. Posterior samples provide predictive probabilities and epistemic uncertainty, while the compact teacher remains available only as an optional triage factor. Across five benchmarks, BLADE remains competitive under a common ranking protocol and reduces adaptive ECE by a macro-average of 60.1% relative to deep ensembles and 78.1% relative to temperature-scaled RotatE. On identical FB15k-237 candidate sets, BLADE also improves ECE, Brier score, and NLL over validation-selected histogram binning and a matched generative ComplEx2 model, with these improvements persisting on a prespecified near-miss pool. Under controlled injected missingness, the full triage score achieves a mean AUC-PR of 0.863, compared with 0.805 for its strongest non-teacher variant. Leakage stress tests show that aligned semantics matter, but they cannot exclude knowledge acquired during LLM pretraining. We therefore claim calibration only for the declared candidate distributions, not for all unobserved triples. --- ### Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks Authors: Zonghao Ying, Jiaqi Yan, Huize Luo, Quanchen Zou, Aishan Liu, Xianglong Liu Published: 2026-08-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.27900v1 Large language models (LLMs) are increasingly deployed in multi-agent systems where a principal agent decomposes tasks and delegates them to subordinate agents that may invoke external tools. Safety alignment, however, is still evaluated almost exclusively under a single-agent threat model, treating safety as a property of the individual LLM. We show that this assumption breaks down: \emph{individual safety alignment fails to transfer to multi-agent settings}. Two failure mechanisms emerge under delegation: \emph{responsibility diffusion} on the principal side and \emph{role-bias compliance} on the subordinate side, jointly converting language-level refusal into actionable harm. We refer to this phenomenon as \textit{delegated misalignment} and study it through a three-condition protocol across 6 frontier LLMs on 49 hazardous tasks. Delegation amplifies end-to-end harm substantially: DeepSeek-V3.2's full-execution rate rises from 30.6\% to 77.6\% once delegation is introduced, and the same model behaves very differently across roles (GPT-5: 22.5\% as a single agent vs.\ 61.2\% as a subordinate). Ablations further show that standard single-layer defenses each fail on their own and can even backfire. We call on the community to move beyond per-model alignment and toward composite safety mechanisms before multi-agent LLM systems are deployed at scale. --- ### Quantifying Protocol-Induced Uncertainty in Comparative Predictive-Model Evaluation: Evidence from Large-Scale Daily PM10 Forecasting Authors: Rafael da Silva, Kiersten Monahan Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.26288v1 Comparative studies of predictive models often end by ranking candidate models, yet these rankings depend on evaluation protocols whose influence is rarely treated as a source of uncertainty. We formalize this problem as protocol-induced ranking uncertainty and introduce a framework that compares ranking displacement caused by switching protocols with displacement produced by conventional choices within a fixed protocol. We quantify these effects using the Protocol Sensitivity Score (PSS) and a full-refit intraprotocol reference. We validate the framework in a large-scale sequential prediction study of daily PM10. Static-split and rolling-origin evaluation are compared across 425 European background stations and 365 US EPA monitors. Switching protocols produces mean PSS values of 0.801 and 0.772 and changes the selected model at 35.3% and 31.5% of stations, respectively. In Europe, intraprotocol perturbations with identical scored targets produce PSS values of 0.072 and 0.230, with winner-swap rates of 0.8% and 4.9%. Between-protocol displacement is therefore substantially larger than the selected within-protocol references. Expanding the candidate set from three to nine models increases the between-protocol winner-swap rate to 60.2% in Europe. The pattern also persists under a frozen protocol applied to held-out background and non-background stations. These results show that model-selection conclusions can depend materially on legitimate evaluation choices. We recommend reporting ranking stability under a small set of defensible intraprotocol perturbations alongside claims of model superiority. --- ### The Architect, the Adversary, and the Judge: Closed-Loop Generation of Standards-Aligned Assessment Items at Scale Authors: Wenhui Chen, Ziyao Lin, Jianlin Chen, Peiji Long, Chi Man Vong Published: 2026-08-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.26087v1 We present CLAIM, a production pipeline for K-12 assessment-item generation coupling a two-stage generate-then-attack protocol (the model drafts as a "curriculum architect", then re-enters the same conversation as a hostile adversarial reviewer), bi-directional few-shot conditioning on accepted and rejected items, the latter carrying the evaluator's diagnosis, and a knowledge dictionary of 44,844 error-correction rules mined from that feedback and retrieved per standard and item type. Across 43,227 scored items over 755 Common Core ELA standards, three item types, and ten LLMs, the pipeline reaches a 97.8% expert-evaluator pass rate on a 9,074-item production run. We then ask what that rate certifies. Re-scoring a stratified sample with three judges from other vendors, blind to the deployed verdict, reproduces the format ordering under every judge and recovers a larger open-set deficit than the deployed evaluator does; but agreement on the accept/reject binary is weak at production prevalence (kappa about 0.13), and the judges agree with each other no better. The level is therefore judge-relative, and with no student-response data our quality evidence is evaluator-judged throughout. The corpus also exposes a robust asymmetry. Multiple-choice and multiple-select generation saturate at 98% or above for both frontier models under a dozen static rules, whereas fill-in-the-blank generation is capability-tiered (82.8-96.7% across five models under a matched rule set, standards, and judge) and plateaus under prompt-only optimization, with error mass shifting between answer-key over-inclusion and omission as rules accumulate. We analyze this as open-set boundary determination, a task autoregressive decoders are structurally ill-equipped to solve, and show the asymmetry recurring when the evaluator itself is distilled: fail-recall rises from 8% to 63% while F1 saturates at 0.25. --- ### What Does Chain-of-Thought Entropy Measure? A Channel Audit of Scaffolding, Routing, and Content Authors: Marios Papamichalis, Regina Ruane Published: 2026-08-26 Categories: stat.ML, cs.IT, cs.LG arXiv: https://arxiv.org/abs/2609.25039v1 Entropy over chain-of-thought tokens decides which tokens receive the policy gradient, which get pruned, and whether a run has collapsed, yet each such statistic reads a next-token distribution mixing three choices: whether to emit connective scaffolding, which connective, and what the substantive continuation should be. Designating a scaffold vocabulary subset separates the three, exactly, for entropy, Kullback--Leibler divergence, and the first-order entropy velocity of a softmax policy. We prove the raw and content conventions disagree about which position is the larger fork on an explicit open region, and bound answer diversity by the content channel plus a leakage term a measured witness certifies. Across twenty-three configurations the scaffold side carries up to 41% of the raw high-entropy set; on a matched-tokenizer ladder, coupling changes only at the math-corpus step while the scaffold's entropy share keeps growing through distillation; a closed-form forecast from one channel correlation tracks selection retention over a 54-point range to five points, unfitted. On compression, the content convention beats raw surprisal in every cell; an answer-leakage audit then corrects our own headline control: re-fed chains earn a quarter to a half of their accuracy from restated answers, and once stripped, no token scorer beats a random contiguous block. --- ### Quantifying Hidden Salt for Precision Healthcare: Sodium Assessment via Joint-Factor Retrieval and Chain-of-Thought Inference Authors: Mingyu Huang, Weiqing Min, Yuehui Fang, Yuna He, Shuqiang Jiang Published: 2026-08-26 Categories: cs.CL, cs.HC, cs.IR, cs.LG arXiv: https://arxiv.org/abs/2609.22171v1 Precision healthcare, particularly for conditions like hypertension and cardiovascular disease, necessitates monitoring of dietary sodium intake. However, tracking this is hindered by the prevalence of hidden salt in cooking, such as sodium in soy sauce and ketchup. While recipes offer a valuable data source for dietary analysis, sodium-rich seasonings are frequently omitted or described ambiguously in instructions. To solve this issue, we propose SALT, a Sodium Assessing & Level Tracking framework adopting an RAG framework to assess sodium content in recipes. Our framework first introduces a Joint-Factor Embedding Retrieval module to locate similar recipes with specified sodium content for addressing the lack of contextual references. These retrieved samples provide contexts for subsequent inference. Then we design a structured 4-hop Chain-of-Thought inference module to refine the vague estimation from language models through a multi-step sodium estimation. To facilitate our study, we further construct a recipe dataset SALT54k with $54,151$ entries labeled with sodium quantities across $11$ common seasonings. Results on SALT54k demonstrate that our method achieves state-of-the-art performance in sodium estimation. Additional real-world validations confirm the effectiveness of our method, demonstrating its potential as a practical solution for AI-assisted precision healthcare. --- ### Multiple latent orderings better predict language model preferences Authors: Aviral Chawla, William H. W. Thompson, Jean-Gabriel Young Published: 2026-08-26 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.22170v1 Language models are frequently employed in settings where they are asked to make value judgments and choices. These observed choices often exhibit intransitivity: A model may prefer item $A$ to $B$ and $B$ to $C$, while also preferring $C$ to $A$. Existing work that models LLM preferences treats such inconsistencies as sampling noise around a single latent ordering. We instead propose that intransitivity reflects the aggregation of multiple latent, internally consistent orderings. We first show that observed inconsistencies cannot be explained by a single ordering under any monotone link function. We then introduce a noise-augmented mixture Bradley-Terry (MBT) model that infers latent preference components from repeated pairwise comparisons. Across seven models and four tasks, a mixture of orderings often explains structural inconsistencies better than single-utility models. We find that aggregate preferences often hide underlying preference heterogeneity. A case study on Moral Machine dilemmas shows that models which disagree on aggregate orderings can still share latent components. Together, these results suggest that LLMs reflect plural preferences. Alignment and evaluation pipelines that treat LLM preferences as a single function, therefore, risk averaging over coherent orderings that different users may endorse differently. --- ### Monocultural Biases: Correlated biases in large language models lead to unequal systemic exclusion rates in hiring Authors: Matthew Bone, Fabian Stephany, Maria del Rio-Chanona Published: 2026-08-26 Categories: cs.CL, cs.CY, econ.GN arXiv: https://arxiv.org/abs/2609.22169v1 Employers are increasingly using large language models (LLMs) to automate their hiring process. This paper investigates the risk of monocultural biases, in which the widespread deployment of large language models homogenizes biases across the labor market, leading to greater systemic exclusion for certain demographic groups. For ten LLMs, we measure hiring biases across their base and post-trained versions to identify which stage, pre-training or post-training, lead to monocultural biases. We find that, compared to their base models, post-trained models are 3.6% less likely to callback older applicants. This negative shift occurs in eight of the ten models that we evaluate. Post-trained models have much more correlated decisions than base models which is likely driven by human capital traits like skills or college major. However, greater consensus among models increases global systemic exclusion rates from 5.6% to 17.3% and exacerbates demographic inequalities, with intersectional systemic exclusion rates ranging from 12.2% to 21.7% for post-trained models. We find that this inequality is primarily driven by age-based discrimination that is exacerbated in post-training. These results indicate that while post-training techniques may improve models' abilities to select the best applicants, they may raise systemic inequality risks for those at the margin by uniformly introducing new biases. --- ### Role-Aware Morgan Fingerprints for Reaction Yield Prediction Authors: Chinmay Mirji, Prashant Shekhar, Foram Madiyar, Hao Peng Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.22167v1 Predicting reaction yield from molecular structure and reaction context can cut experimental trial-and-error and speed up condition screening in synthetic chemistry. Recent methods for this task use learned representations such as graph neural networks or Transformer encoders over reaction SMILES (Simplified Molecular Input Line Entry System), but these approaches carry heavy preprocessing overhead and can break when input formatting is inconsistent. We propose MFP, a reaction yield prediction method built on role-aware Morgan fingerprints where count-based circular fingerprints are computed for each reaction component, aggregated by chemical role (reactant, reagent, product), and combined with transformation-sensitive difference features into a fixed-length reaction descriptor fed to a feed-forward neural regressor. We test MFP against state of the art methods such as YieldBERT (with and without data augmentation) and GNAN (graph neural network) on the Suzuki-Miyaura and Buchwald-Hartwig benchmarks using a shared preprocessing and evaluation protocol. MFP reaches R2 = 0.878 on Suzuki-Miyaura and R2 = 0.969 on Buchwald-Hartwig while training an order of magnitude faster than graph- or Transformer-based alternatives. A formal complexity analysis confirms that MFP folds all representation cost into a one-time preprocessing step, removing the per-epoch message-passing overhead that graph methods carry. An ablation over fingerprint radius and folded vector length shows that radius-2 representations at nBits =2048 give the best balance of accuracy, speed, and cross-split stability on both datasets. These results establish MFP as an effective, reproducible, and efficient baseline for reaction yield prediction. --- ### Clustering-Based Collective Anomaly Detection in IoT Systems: A Graph Neural Network Approach Authors: Dalila Khettaf, Djamel Djenouri, Zeinab Rezaeifar, Youcef Djenouri Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.22166v1 The rapid advancement of Internet of Things (IoT) technology has led to the widespread deployment of smart, interconnected devices across a range of domains. However, this expansion has also resulted in a substantial increase in network traffic, creating more opportunities for malicious actors to launch cyberattacks and compromise sensitive information, thereby increasing the need for effective anomaly detection. The state-of-the-art in anomaly detection has predominantly focused on point anomalies. In contrast, the detection of collective anomalies remains relatively under-explored in the literature. In this paper, we introduce Unsupervised Graph Collective Anomaly Detection (UGCAD), a novel frame- work designed to identify collective anomalies in IoT network traffic. Unlike many existing methods, UGCAD operates on graph-structured data without any prior knowledge of group labels or membership. It leverages a variational graph autoencoder (VGAE) to learn the graph representation, which is subsequently used to enhance a clustering algorithm for effective grouping of nodes. To detect collective anomalies, clusters identified as normal are first aggregated and refined, after which anomaly scores are applied to detect collective anomalies. Extensive experiments conducted on the CICIoT2023 and ToN-IoT network datasets demonstrate the effectiveness of UGCAD in both clustering and collective anomaly detection (CAD). Furthermore, comparative evaluations against several traditional and state-of-the-art clustering-based CAD approaches confirm the superiority of UGCAD in accurately detecting collective anomalies. --- ### TARGet: Topology-Aware Fusion-based Radio Frequency Circuit Functional Modeling using Graph Neural Networks Authors: Soroosh Noorzad, Sebastian Bodero, Morteza Fayazi Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.22165v1 Automatic synthesis of analog and Radio Frequency (RF) circuits is an emerging area that requires an efficient circuit modeling method. In recent years, Machine Learning (ML) solutions have played a promising role in this regard. However, many existing ML approaches require separate training data for each circuit topology, even when a single circuit component is added or removed. In addition, they overlook circuit topology information, which limits their ability to capture complex component interactions. Furthermore, they rely on fully connected neural networks with flat feature representations, which require substantial amounts of training data. In this work, we propose an open-source topology-aware RF circuit modeling method, TARGet. Our model considers the circuit at two levels: sub-circuits and the overall circuit topology. At the sub-circuit level, TARGet leverages S-parameter representations to capture sub-circuit behavior rather than relying on individual circuit components, providing a reusable behavioral abstraction for RF building blocks. Moreover, TARGet explicitly incorporates circuit topology information into the model, enabling it to learn across multiple topologies. TARGet introduces a novel fusion-based architecture that integrates Graph Neural Networks (GNNs) and sub-circuit connectivity-aware neural networks to improve data efficiency. Experimental evaluation across multiple RF circuit topologies demonstrates that TARGet achieves sub-1% prediction error while reducing the required training data by up to 35.5x compared to state-of-the-art (SOTA) approaches. Furthermore, TARGet achieves 9.7x higher prediction accuracy under a strict 1% error threshold relative to SOTA models. A held-out matching-network evaluation further demonstrates zero-shot transfer to an unseen sub-circuit topology, where TARGet reduces NMAE by up to 45%. --- ### A Multi-Agent Pipeline for Source-Grounded Synthetic Note Generation from Longitudinal Structured EHR Authors: Nina Fatehi, Reihaneh Hassanzadeh, Meysam Ghaffari, Animesh Agarwal, Carlos Morato Published: 2026-08-26 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.22164v1 Structured EHR is abundant but sparse, coded, and difficult to use directly for note-centric clinical modeling. We present MedNotes, a multi-agent synthetic data generation pipeline that converts longitudinal structured EHR into source-grounded clinical note representations under explicit quality control. MedNotes treats structured-data-to-text synthesis as a closed-loop agentic process: a generator proposes a note, evaluator agents diagnose factual, coverage, structural, and hallucination-related failures, and a router accepts, revises, or rejects the draft. On 1,485 EHRSHOT encounters, MedNotes achieves a 91.4% pass rate, with mean factual accuracy of 0.980, completeness of 99.1%, structural fidelity of 0.761, and 0.028 critical hallucinations per encounter. Iterative refinement improves acceptance from 69.4% to 91.4%. The resulting synthetic corpus improves downstream CPT prediction and paragraph-level section prediction when combined with limited real data. --- ### MechaTerp-TRACE: A Novel Approach for Component Ablation Analysis in Language Models Authors: Brandon Colelough, Davis Bartels, Madeline Bittner, Dina Demner-Fushman Published: 2026-08-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.22163v1 Interpretability research on large language models has produced accounts of factual recall in feed-forward layers and of token relationships in self-attention, but little work offers a unified way to compare the causal contribution of different architecture components to a model's output. We introduce MechaTerp (the Mechanistic Interpretability suite) -TRACE (subset for Teacher-forced Registry of Ablated Component Effects), an architecture and study that measures how much each registered component of a language model supports the production of a named entity. TRACE ablates one component at a time and measures the resulting change in the output distribution at a fixed answer token, so component types from whole transformer blocks down to individual neurons and output logits can be compared on a common scale. We apply it to thirteen instruction-tuned dense decoder models spanning five families and one to thirty billion parameters, ablating 49,656 components across 48 medical and 42 general-knowledge prompts. We find that the components carrying the most effect are the same few, positionally fixed components in every model, regardless of which entity a prompt asks about, and that once these are removed, the remaining support is close to evenly spread in eleven of the thirteen models. Apparent localisation of entity knowledge is therefore largely attributable to generic generation machinery, which has direct consequences for methods that assume entity knowledge sits in a findable place, including targeted knowledge editing. --- ### Beyond Raw Context Transfer: Representation-based Federated Retrieval-Augmented Generation Authors: Can Peng, Yu Liu, Yingyu Yang, Anjie Le, Yuyuan Liu, Qianye Yang, J. Alison Noble Published: 2026-08-26 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.22162v1 Retrieval-augmented generation (RAG) improves the factuality of large language models (LLMs) and vision-language models (VLMs) by grounding generation in external knowledge. However, most existing RAG frameworks assume a centralized retrieval corpus, which is often impractical in sensitive domains such as healthcare, where data are inherently distributed and raw content cannot be directly shared across institutions. Recent efforts on decentralized RAG primarily follow prompt-based paradigms that exchange raw, human-readable retrieved content, leading to substantial inference-time computational overhead and direct exposure of retrieved information. To address these limitations, we propose Representation-based Federated RAG (FedRepRAG), a decentralized RAG framework that keeps raw documents at their owning clients and exchanges only compact latent representations during cross-client retrieval. To integrate retrieved knowledge, we introduce a collaboratively trained projector that converts retrieval embeddings into generator-compatible representation tokens for a frozen LLM/VLM backbone. Experiments across decentralized visual question answering (VQA) and question answering (QA) benchmarks show that FedRepRAG consistently outperforms direct inference and local retrieval baselines while substantially reducing retrieval-context length and inference-time computational overhead compared with raw-context transfer. Further analyses confirm the importance of query-relevant retrieved representations and characterize the residual representation-level leakage associated with representation exchange. Overall, FedRepRAG provides an effective and efficient framework for federated RAG without transferring raw retrieved content. --- ### Didactic knowledge or Clinical Cases? How Data Types Shape Medical Large Language Models Authors: Yuzheng Fan, Haochun Wang, Sendong Zhao, Xiao Han, Ming Ma, Bing Qin Published: 2026-08-26 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.22161v1 Medical large language models are commonly trained on mixtures of didactic data (e.g., textbooks) and clinical data (e.g., patient records), yet how these data types differentially shape model capabilities remains unclear. We address this issue with token-matched experiments that vary the didactic-to-clinical ratio and analyze how data composition affects performance, capability profiles, and error patterns across knowledge-intensive and clinic-oriented tasks. We uncover an asymmetric transfer across task types: clinical data improves clinic-oriented tasks while remaining competitive on knowledge-intensive ones, whereas didactic data mainly improves knowledge-intensive tasks. Error analysis suggests a knowing-doing gap, where improvements in knowledge recall do not reliably generalize to clinical reasoning. We further observe that modest amounts of clinical data yield most of the gains on EHR-grounded tasks, while the optimal mixture ratio varies with the knowledge and clinical reasoning demands of downstream tasks. These findings suggest that medical LLM data curation should be application-driven, with higher proportions of clinical data preferred for reasoning-intensive use cases. --- ### Uncertainty and Business-Aware Remaining Useful Life Estimation for Semiconductor Manufacturing Authors: Davide Frizzo, Francesco Borsatti, Gian Antonio Susto Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.22160v1 Semiconductor manufacturing relies on tightly interconnected components, so early identification of the assets most likely to fail is essential to prevent a single breakdown from disrupting the entire production pipeline. Maintenance planning must therefore balance unexpected failures against prematurely interrupted operating life. We present a Predictive Maintenance (PdM) framework combining Deep Learning (DL) sequence models and Simoultaneous Quantile Regression (SQR) for uncertainty-aware Remaining Useful Life (RUL) estimation and risk-aware maintenance decisions. Several architectures are compared on ion-milling data from the 2018 PHM Data Challenge (PHM18), including architectures based on State Space Models (SSM), using prediction and business metrics: Unexpected Breaks (UB), Unexploited Lifetime (UL), and a cost-weighted objective. Diagonal State Spaces (S4D) delivers the best Remaining Useful Life (RUL) estimates across quantiles and, relative to Preventive Maintenance (PvM) baselines, substantially lowers business cost by avoiding systematically early interventions. The results support uncertainty-aware, cost-sensitive maintenance planning in semiconductor production. --- ### StepKV: Step-Aware KV Cache Compression for LLM Agents Authors: Boyu Feng, Jiahong Liu, Yifan Li, Wenhao Yu, Zexuan Qiu, Yuliang Sun, Ming Shen, Xiang Li, Quanyu Dai, Irwin King Published: 2026-08-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.22158v2 Key-value (KV) caching is essential for efficient autoregressive large language model (LLM) inference, but the cache grows linearly with context length, increasing storage and decoding costs. KV cache compression mitigates this cost by retaining only a subset of cached tokens. This challenge is particularly important for multi-step LLM agents, where a query expands into trajectories of reasoning, tool interactions, and retrieved observations. Existing pruning methods typically treat the cache as a flat token stream and rank tokens by recency or attention saliency. This creates a mismatch between the unit of compression and the unit of reasoning: token-level pruning removes individual entries, whereas useful information in multi-step agents is often organized into reasoning steps with uneven and delayed importance. Consequently, an early observation or intermediate decision may receive little recent attention yet remain essential for later evidence synthesis. We term this failure mode Reasoning Continuity Disruption.These observations motivate KV cache compression that jointly considers token- and reasoning-step-level information. StepKV addresses this goal by treating reasoning steps as first-class retention units. It associates cache entries with their generating steps, estimates step utility from trajectory-derived signals, and combines this utility with token-level saliency. The resulting scores globally rank prunable tokens, from which StepKV retains the top-scoring entries under a target budget. StepKV thus provides a step-centric perspective for agent KV cache compression. Across multi-hop QA and long-horizon web reasoning tasks, StepKV sustains accuracy under low KV budgets where token-level baselines degrade sharply, offering a more robust efficiency-accuracy trade-off for multi-step agent inference. --- ### PAGE: Partition-Aware Gated KV-Cache Eviction Authors: Pankaj Kumar, Subhankar Mishra Published: 2026-08-26 Categories: cs.LG, cs.CL, cs.CR arXiv: https://arxiv.org/abs/2609.22157v2 KV-cache eviction can do more than compress. In long-context LLMs, keeping only some cached tokens sometimes matches or exceeds full-cache accuracy, because many redundant prefill tokens otherwise dilute attention away from the tokens that carry the answer. This benefit is not uniform, and evicting the wrong tokens can drop accuracy to zero on tasks that require precise retrieval, so the useful question is not only which tokens to keep but also whether to evict this input at all. We show that one label-free number computed from the prefill attention, the drop between early and late layers in how much attention heads agree on which tokens to read, predicts per input, before any decoding, which of the two cases an input falls under. We build this into PAGE (Partition-Aware Gated Eviction), a wrapper that runs any SnapKV-style evictor when the drop is large and keeps the full cache when it is small, with no training, labels, or fine-tuning. PAGE is a safety mechanism rather than a compressor, so we measure it by the failures it prevents. It cuts the harm rate on capacity-bound inputs from 0.75 to 0.026, and on multi-key retrieval with Mistral-7B plain SnapKV falls from 99\% to 0\% as the budget shrinks, while PAGE holds it at 89\%. Elsewhere, it passes the base evictor through unchanged, which is the intended behaviour and is what we observe in 8 of 16 cells. Code is available at https://anonymous.4open.science/r/PAGE-018239. ---