Academic Intelligence · Curated Daily

Explore the Frontier of Global Academia

AcademicHub aggregates real-time literature from top journals and preprint platforms. Build your personal research radar and let large language models compile cross-disciplinary analysis briefings automatically.

01.
arXiv (CS.CV) 2026-06-17

Structured Adversarial Camouflage via Voronoi Diagrams

Pixel-wise adversarial patches are computationally heavy and often visually detectable, limiting utility in security-critical systems. We present adversarial Voronoi camouflage that optimizes only seed-point locations under fixed, printable palettes using a soft assignment, producing structured, splinter camouflage-like patterns without additional regularization. Evaluated on person detection with COCO-style AP@[.5:.95], naive placement (Inria -> COCO) performs comparably bad, while garment-level application via segmentation mask (3DPeople) results in a significant AP drop. The attack transfers to out-of-domain backgrounds and across detector families (YOLOv9/10/11/12), indicating robustness in black-box settings. Repainting with different palettes largely nullifies the effect, and single-color tweaks show limited tolerance (

02.
arXiv (CS.CL) 2026-06-11

Reassessing High-Performing LLMs on Polish Medical Exams: True Competence or Bias-Driven Performance?

Large language models (LLMs) in medicine are mainly evaluated using multiple-choice question answering (MCQA), which can overestimate real clinical ability due to guessing strategies and answer biases. To address these limitations, we introduce an expanded and more challenging benchmark based on Polish medical exams, adding over 15,000 questions, two new domains, and four structural modifications that reduce MCQA-specific artifacts and better test reasoning. We evaluate 21 LLMs and show that evaluation design strongly affects results. Under our harder setup, the best model (Qwen3.5-122B) drops by 28.4 and 31 pp on English and Polish exams, respectively. Despite low evidence of data contamination, standard MCQA scores do not reliably reflect true medical competence. To facilitate further research, we make our benchmark publicly available.

03.
arXiv (CS.CV) 2026-06-15

Value-order Decomposition for Generalist Anomaly Detection

Industrial anomaly detection suffers from limited data, making cross-domain generalization particularly challenging. Generalist Anomaly Detection (GAD) aims to train a unified model on a source domain that can effectively detect anomalies in unseen target domains. In the initial semantic feature space, strong entanglement between anomalies and object categories or defect types hinders effective generalization across domains. Recent works address this issue by projecting features into a residual space; however, such methods primarily increase cross-domain overlap for normal features, while anomalous features remain specific to object categories, defect types and data domains, leading to poor alignment and generalization. To address this limitation, we propose Value-order Decomposition (VOD), a simple yet effective technique that bridges three types of generalization gaps across object categories, defect types (including real and synthetic defects), and data domains. VOD disentangles and suppresses object-category-, defect-type-, and domain-specific information, promoting alignment within normal and abnormal samples while preserving their separability, thereby enabling robust generalization across the three gaps. Leveraging the strong alignment between real and synthetic defects within the same object, we perform anomaly detection using only normal and synthetic-abnormal reference, and effectively generalize to unseen real defect types. Experiments on diverse industrial and medical benchmarks demonstrate that our method, using a simple cut-and-paste anomaly simulation strategy, achieves strong generalization across the three gaps.

04.
arXiv (CS.CL) 2026-06-18

GateMem: Benchmarking Memory Governance in Multi-Principal Shared-Memory Agents

Memory benchmarks for LLM agents largely assume single-user settings, leaving shared assistants for hospitals, workplaces, campuses, and households understudied. In these deployments, multiple principals write to a common memory pool and query it under different roles, scopes, and relationships, so memory quality requires governance as well as recall. We introduce GateMem, a benchmark for multi-principal shared-memory agents. GateMem jointly evaluates utility for legitimate long-horizon requests with state updates, access control across contextual authorization boundaries, and agent-facing active forgetting after explicit deletion requests. It spans medical, office, education, and household domains, with long-form multi-party episodes, incremental memory injection, hidden checkpoints, structured judging, and leak-target annotations. Across diverse baselines and backbone models, no method simultaneously achieves strong utility, robust access control, and reliable forgetting. Long-context prompting often yields the best governance score at high token cost, while retrieval-based and external-memory methods reduce cost yet still leak unauthorized or deleted information. These results show current memory agents remain far from reliable shared institutional deployment.

06.
arXiv (CS.CL) 2026-06-18

Breaking the Solver Bottleneck: Training Task Generators at the Learnable Frontier

The limiting resource for training agents via reinforcement learning (RL) is increasingly frontier task supply: valid, solvable tasks just difficult enough to train the current model. As reasoning and agentic models improve, fixed task distributions saturate, while naive synthetic generation yields tasks that are trivial, impossible, or ill-posed. Training a task generator with RL to optimize validity and learnability can address this bottleneck, but direct optimization requires repeated solver rollouts per candidate. For software-engineering (SWE) tasks, a single rollout can take tens of minutes; solver-in-the-loop generator training is intractable. We introduce PROPEL, a solver-amortized framework for training task generators at the targeted solve rate. PROPEL trains a lightweight activation probe on a one-time labeled corpus of generated tasks and solver outcomes. The probe predicts target-solver pass rate from a frozen generator reference model and serves as a proxy for solve rate during generator optimization, reducing generator evaluation to a single forward pass. Across math, code, and software-engineering at multiple model scales, PROPEL shifts generation toward the targeted solve rate: for coding, tasks generated at the learnable frontier increase from $10.1\% \rightarrow 20.0\%$ for a Qwen2.5-3B-Instruct solver and from $5.3\% \rightarrow 12.6\%$ for a Qwen2.5-7B-Instruct solver. For SWE, PROPEL increases the share of generations at the targeted solve rate from $9.8\% \rightarrow 19.6\%$ for Qwen3.5-27B on repositories not seen during training of probe and generator.

07.
arXiv (CS.CL) 2026-06-18

Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States

Progress in legal AI increasingly depends on access to authoritative legal text at scale. Yet one of the most consequential layers of American law remains largely absent from existing machine-readable corpora: local ordinances. Local codes govern zoning, housing, business licensing, public health, noise, animal control, and many other domains of everyday regulation, but they are fragmented across vendor platforms designed for human browsing rather than bulk research access. We introduce LOCUS - the Local Ordinance Corpus for the United States - a comprehensive corpus and county-harmonized access layer for U.S. municipal and county ordinance codes. The raw corpus, available for release to researchers, represents nearly all publicly available municipal and county ordinance codes. The resulting raw corpus contains codes from 9,239 cities and counties. A smaller county-harmonized LOCUS access layer provides coverage for the largest 2,309 of 3,144 U.S. counties, accounting for a majority of the population. We use OCR to handle the myriad of document formats that have kept the law from being a public resource. We release the corpus with coverage metadata to support reproducibility, downstream legal AI research, and the incremental expansion of machine-readable access to local law. We train a collection of ModernBERT-based classifiers and scorers to facilitate analyzing U.S. local law among several dimensions, such as opacity and paternalism, that have not previously been studied at this scale. LOCUS-v1 and its derivative models are available at: https://huggingface.co/datasets/LocalLaws/LOCUS-v1

08.
arXiv (CS.LG) 2026-06-17

Domain-Validity-Gated Metamorphic Testing of Scientific ML Surrogates

arXiv:2606.17529v1 Announce Type: cross Abstract: Scientific machine-learning (SciML) surrogates approximate expensive simulations, but exact expected outputs for arbitrary inputs are unavailable (the oracle problem). Metamorphic testing checks relations across executions, yet a candidate relation is not automatically valid: its preconditions, output mapping, and the numerical floor of the scoring operator determine whether a violation is meaningful. We study how candidate metamorphic relations (MRs) can be screened for domain validity and turned into executable, oracle-free test assets for SciML surrogates. We propose (i) a domain-validity rubric that admits a candidate only when its tolerance dominates the operator's numerical floor and its preconditions hold; (ii) an MR-card executable-asset format recording source cases, transformations, metrics, tolerances, and typed relation-level verdicts; and (iii) a case-study protocol on MeshGraphNets cylinder-flow surrogates, with a claim ledger binding every result to a tracked artifact. On a MeshGraphNets checkpoint, node permutation holds to machine precision, mirror-y is a bounded out-of-distribution stress finding rather than an exact symmetry, and absolute conservation stays deferred while a reference-relative guard passes. The same readings hold across held-out trajectories, a checkpoint roster, three further architectures, and PhysicsNeMo. On a second CFD task (compressible airfoil) the predicate instead rejects incompressible continuity on physical grounds, showing it reasons about domain validity rather than running a fixed checklist. On a second PDE family, FNO Burgers and heat surrogates run full admit/reject/execute verdicts. The evidence spans two CFD tasks and a second PDE family, supporting a validity-aware bridge from candidate MRs to auditable SciML test assets that separates model-level violations from out-of-domain applications.

09.
arXiv (CS.CV) 2026-06-19

Smol-GS: Compact Representations for Abstract 3D Gaussian Splatting

We present Smol-GS, a novel method for learning compact representations for 3D Gaussian Splatting (3DGS). Our approach learns highly efficient splat-wise features to model 3D space, which capture abstracted cues, including color, opacity, transformation, and material properties. We propose octree-derived positional encoding, which explicitly models spatial locality and enhances representation efficiency. We further apply entropy-based compression to exploit feature redundancy and compress splat coordinates using a recursive voxel hierarchy. This design enables orders-of-magnitude reduction in storage while preserving representation flexibility. Smol-GS achieves state-of-the-art compression performance on standard benchmarks with high-level rendering quality.

10.
arXiv (CS.LG) 2026-06-12

Authority, Truth, and Citation Bias: A Large-Scale Multi-Domain Benchmark for Studying Epistemic Susceptibility in Large Language Models

arXiv:2606.13104v1 Announce Type: new Abstract: Large language models are increasingly deployed in citation-augmented settings, yet the effect of citation presence on model behavior independent of factual content remains poorly understood. We introduce AuthorityBench, a 220,564-prompt multi-domain benchmark that isolates how citation-based authority signals influence epistemic behavior in LLMs. The benchmark uses a fully balanced 2x2 factorial design crossing claim veracity with citation veracity, the first to do so, across four domains (general knowledge, science, law, and medicine), with controlled variation over 40 prompt templates, four venue prestige tiers, and a country-coded author name dataset. Evaluating seven models on 12 structured research questions, we find that citation presence, whether real or fabricated, consistently increases hallucination rates relative to a no-citation baseline. The effect is strongest when fabricated citations accompany true claims, raising hallucination rates by 3 to 22 percentage points and reaching 35 to 77% in the general knowledge domain, while legal claims are comparatively robust and venue prestige and author demographics show negligible impact. All datasets and evaluation code are available at: https://github.com/floating-reeds/AuthorityBench

11.
arXiv (CS.CL) 2026-06-16

Not What, But How: A Framework for Auditing LLM Responses across Positioning, Generalization, Anthropomorphism, and Maxims

Large language models (LLMs) are being increasingly used to answer subjective, information-seeking questions, where users are sensitive to how responses are communicated, not just whether the answers are correct. Existing LLM evaluations for subjective cultural queries largely focus on factual correctness, ignoring how the response is framed. To this end, we introduce FRANZ, an automated FRAmework for respoNse characteriZation to conduct communicative audit of LLM responses along four dimensions: cultural positioning, use of generalizing language, anthropomorphic cues, and adherence to conversational maxims. To enable this evaluation, we contribute SQUARE - a corpus of 376k subjective questions sourced from 57 subreddits, and mapped to 7 countries and 19 question categories. We demonstrate FRANZ's applicability by scoring responses from three open-weight LLMs. We observe that LLMs show statistically significant differences in the frequency with which they employ each response characteristic. Unlike single-dimensional audits, FRANZ reveals that insider positioning and anthropomorphism are positively coupled, with the degree of coupling varying by country, providing a diagnostic lens for identifying framing divergences.

12.
medRxiv (Medicine) 2026-06-12

Genetic basis of dynamic brain states reveals cellular and disease associations

Dynamic resting-state fMRI captures the time-varying patterns of brain activity that are obscured by static approaches. Hidden Markov Models (HMMs) characterise these dynamics as recurring whole-brain states and quantify their fractional occupancy (FO), the proportion of time spent in each state, yet the biological basis of inter-individual variation in FO remains unclear. Using data from 52,335 White UK Biobank participants, with replication in East and South Asian subsamples, this study examined the heritability, cellular and neurotransmitter basis of brain states, and their links with complex phenotypes. FO was significantly heritable and enriched for neuronal populations, particularly glutamatergic and GABAergic signalling. Analyses identified shared and state-specific loci and revealed genetic correlations, colocalisation, and potential causal relationships between FO and several phenotypes, including educational attainment, sleep duration, and disease risk. These findings establish dynamic brain states as biologically grounded intermediate phenotypes, linking genetic variation to neural dynamics, diseases and traits.

13.
arXiv (CS.AI) 2026-06-12

Decentralized Autoregressive Generation

arXiv:2601.03184v3 Announce Type: replace-cross Abstract: The decentralization of autoregressive generation has attracted considerable attention in recent years as a solution to scaling bottlenecks. However, despite promising empirical results, this paradigm currently lacks rigorous theoretical justification. In this work, we formally establish the theoretical equivalence between decentralized and centralized training. To achieve this, we adapt the Discrete Flow Matching framework for autoregressive generation, leveraging its inherent properties to demonstrate that global models naturally decompose into independent experts. Finally, we conduct extensive experiments across diverse multimodal benchmarks, empirically validating that decentralized training maintains competitive parity with standard centralized architectures.

14.
arXiv (quant-ph) 2026-06-16

Instrument-based quantum resources: quantification, hierarchies and towards constructing resource theories

arXiv:2508.09134v3 Announce Type: replace Abstract: Quantum resources are certain features of the quantum world that provide advantages in certain information-theoretic, thermodynamic, or other useful operational tasks that are outside the realm of what classical theories can achieve. Quantum resource theories provide us with an elegant framework for studying these resources quantitatively and rigorously. While numerous state-based quantum resource theories have already been investigated, and to some extent, measurement-based resource theories have also been explored, instrument-based resource theories remain largely unexplored, with only a few notable exceptions. As quantum instruments are devices that provide both the classical outcomes of induced measurements and the post-measurement quantum states, they are quite important, especially for scenarios where multiple parties sequentially act on a quantum system. In this work, we study several instrument-based resource theories, namely (1) the resource theory of information preservability, (2) the resource theory of (strong) entanglement preservability, (3) the resource theory of (strong) incompatibility preservability, (4) the resource theory of traditional incompatibility, and (5) the resource theory of parallel incompatibility. Furthermore, we outline the hierarchies of these instrument-based resources and provide measures to quantify them. We then also established a relationship between our resource measure and the advantage in an information-theoretic task. In short, we provide a detailed framework for a wide variety of instrument-based quantum resource theories.

15.
arXiv (quant-ph) 2026-06-17

Unclonable Encryption in the Haar Random Oracle Model

arXiv:2603.11437v2 Announce Type: replace-cross Abstract: We construct unclonable encryption (UE) in the Haar random oracle model, where all parties have query access to $U,U^\dagger,U^*,U^T$ for a Haar random unitary $U$. Our scheme satisfies the standard notion of unclonable indistinguishability security, supports reuse of the secret key, and can encrypt arbitrary-length messages. That is, we give the first evidence that (reusable) UE, which requires computational assumptions, exists in "microcrypt", a world where one-way functions may not exist. As one of our central technical contributions, we build on the recently introduced path recording framework to prove a natural ``unitary reprogramming lemma'', which may be of independent interest.

16.
arXiv (CS.AI) 2026-06-15

Hybrid Open-Ended Tri-Evolution Makes Better Deep Researcher

arXiv:2606.13710v1 Announce Type: new Abstract: Deep research and agent evolution serve as de-facto tasks for AI agents in real-world applications toward artificial general intelligence. The former enables autonomous retrieval and integration of information in open-ended environments to tackle open-ended research tasks, yet it is constrained by the static parametric deep research capabilities of agent systems. The latter allows agents to autonomously interact with the environment to gain experiences that evolve model capabilities. However, its effectiveness has been widely validated only on verifiable tasks with standard answers, leaving a gap with open-ended research tasks. To bridge these two critical tasks, we propose the Hybrid Open-Ended Tri-Evolution (HOTE) framework, which leverages hybrid-mode reinforcement learning to facilitate the collaborative evolution of a proposer, solver and judge based on web-scale knowledge, moving toward autonomous evolving agents in open-ended tasks and environments. Extensive experiments on three long-form deep research benchmarks demonstrate that the 8B model trained via HOTE surpasses the strongest static open 8-32B models as well as those trained by state-of-the-art deep research training methods with less time overhead, and further verify that the evolution of all three modules in HOTE is indispensable.

17.
arXiv (CS.CV) 2026-06-11

CFCamo: A Counterfactual Detect-or-Abstain Framework for Camouflaged Object Detection

Vision-language reinforcement learning has recently shown strong target-present localization for camouflaged object detection (COD). Yet localization is only one side of the decision: when the agent faces an ordinary image with no camouflaged target, will it still claim that a camouflaged object exists? Standard COD training and evaluation data are positive-only, so agents optimized under this setting can acquire an over-detect bias, a task-specific form of object hallucination that standard COD evaluation leaves unmeasured. To quantify this target-absent behavior, we construct Counterfactual COD (CF-COD), a paired benchmark that removes the camouflaged target from each held-out COD evaluation image while preserving a plausible background. CF-COD evaluates whether a model detects the target on the original image and abstains on the target-absent counterfactual, summarized by Pair Accuracy (PA). We further introduce CFCamo, a paired counterfactual framework for COD with abstention. For training, CFCamo optimizes a Qwen3-VL-4B-Instruct agent with Counterfactual Sequence Policy Optimization (CSPO), which samples paired original-counterfactual rollouts and uses a Counterfactual Paired Reward (CPR) to couple original-image detection with counterfactual abstention. On CAMO-test, CFCamo improves S_alpha by +3.7 pp over the prior RL-based COD baseline; across CF-COD, it reaches 80.0-90.8% PA. Ablations show that removing counterfactual coupling reduces PA to 1.4-5.2% despite strong target-present COD scores, showing that target-present evaluation alone does not characterize detect-or-abstain behavior. Overall, these results indicate that CFCamo improves COD agents by coupling target-present detection with target-absent abstention, rather than merely strengthening target-present localization. Code and data are available at https://github.com/suhang2000/CFCamo.

18.
arXiv (quant-ph) 2026-06-16

Quantum Illumination with Symmetry-Constrained Random Unitaries

arXiv:2606.15586v1 Announce Type: new Abstract: Quantum illumination provides a quantum advantage in detecting weakly reflecting objects embedded in a noisy environment, even when environmental noise destroys most of the initial entanglement. We investigate this advantage using Haar-random probe states constrained to symmetry-resolved subspaces. Employing tools from quantum channel discrimination and asymptotic hypothesis testing, we derive the discrimination exponents associated with Haar-random probe ensembles and identify the role of symmetry in determining their performance. We show that typical states drawn from fixed-charge sectors achieve the same asymptotic quantum-illumination advantage as maximally entangled probes. In particular, we show that the effective thermal-noise suppression and the corresponding Chernoff exponent are governed by the dimension of the accessible symmetry sector. Our results reveal that the operational resource underlying quantum illumination can be generalized from fine-tuned structure of a specific probe state to the existence of a large symmetry-protected correlation subspace. These findings establish a direct connection between quantum illumination, symmetry-resolved typicality, and quantum channel discrimination, and demonstrate that near-optimal quantum hypothesis testing resources can emerge naturally from generic many-body quantum states constrained by conservation laws.

19.
arXiv (CS.CV) 2026-06-11

Continual Learning with Support Boundary Experience Blending

Continual learning (CL) seeks to mitigate catastrophic forgetting when models are trained with sequential tasks. A common approach, experience replay (ER), stores past exemplars but only sparsely approximates the data distribution, yielding fragile and oversimplified decision boundaries. We address this limitation by introducing Support Boundary Data (SBD), generated via differential-privacy-inspired noise into latent features to create boundary-adjacent representations that implicitly regularize decision boundaries. Building on this idea, we propose Experience Blending (EB), a framework that jointly trains on exemplars and SBD through a dual-model aggregation strategy. EB has two components: (1) latent-space noise injection to generate support boundary data, and (2) end-to-end training that jointly leverages exemplars and SBD. Unlike standard experience replay, SBD enriches the feature space near decision boundaries, leading to more stable and robust continual learning. Extensive experiments on CIFAR-10, CIFAR-100, Tiny ImageNet, and ImageNet1K demonstrate consistent accuracy improvements of 10%, 6%, 13%, 2%, respectively.

20.
arXiv (CS.LG) 2026-06-12

Fed-FBD: Federated Functional Block Diversification for Isolation, Privacy, and Surgical Unlearning

arXiv:2606.12679v1 Announce Type: new Abstract: Federated learning (FL) enables collaborative model training without sharing raw patient data, but standard approaches such as FedAvg treat each client as a black box and provide no mechanism for isolating an adversarial contributor, auditing per-client influence, or honoring a departed participant's right to be forgotten. We present Fed-FBD (Federated Functional Block Diversification), a modular federated architecture that decomposes a ResNet backbone into six functional blocks (the stem, four residual groups, and the classification head) and maintains a warehouse of N color variants, each assembled from independently tracked and contributor-stamped blocks. Fed-FBD provides three capabilities absent in FedAvg: (i) architecturally guaranteed block-level isolation, so that an adversarial or mislabelled client cannot contaminate the clean colous; (ii) privacy-by-design, where membership inference advantage is already indistinguishable from chance before any privacy mechanism is applied; and (iii) surgical machine unlearning of a departed participant's contribution at sub-second cost and without retraining. Experiments on six MedMNIST-2D datasets, PathMNIST at 224x224, and CIFAR-10 show that Fed-FBD trades a modest 0.3%-3.1% IID accuracy gap on the adequately sized datasets for these guarantees, remains within 0.8%-4.0% of FedAvg at Dirichlet alpha=1.0 on three of four datasets, and confines all six adversarial attacks we study to the poisoned client's own blocks with at most +/-0.01 AUC drift on the clean colors.

21.
arXiv (CS.LG) 2026-06-11

DeMix: Debugging Training Data with Mixed Data Error Types by Investigating Influence Vectors

arXiv:2606.11616v1 Announce Type: new Abstract: High-quality training data is essential for the success of machine learning models. However, real-world datasets often contain mixed types of errors arising from systematic flaws in data preparation pipelines, including label errors, feature errors, and spurious correlations. Effective debugging of training data requires both detecting erroneous samples and identifying their specific error types to enable targeted repair, yet existing data cleaning and attribution methods fail to adequately address this dual requirement. In this paper, we propose DeMix, a novel framework that simultaneously diagnoses erroneous samples and their error types. Our key insight is that different error types produce distinct patterns on model behavior. DeMix captures such error-specific patterns by influence vectors that characterize how each training sample affects model predictions across all validation samples. We formulate training data debugging as a multi-label classification problem where a classifier is developed to predict error types directly from influence vectors. We further introduce an intervention-based learning strategy that guides the classifier to capture invariant rationales specific to each error type, ensuring the learned classifier generalizes effectively. Empirical evaluations on 11 tasks across tabular data prediction, recommendation systems, and LLM alignment demonstrate that DeMix significantly outperforms state-of-the-art approaches, achieving a 22.61% improvement in data debugging F1-score and a 9.32% gain in task model performance after data repair. Code is available at: https://github.com/SJTU-DMTai/DeMix.

22.
arXiv (CS.LG) 2026-06-17

A Closer Look at Failure Modes in Temporal Understanding of Large Audio-Language Models

arXiv:2606.17417v1 Announce Type: cross Abstract: Large Audio Language Models (LALMs) achieve strong performance on a variety of audio understanding tasks but continue to struggle with temporal reasoning, a fundamental capability central to human auditory perception. Understanding the causes of these failures remains challenging as existing benchmarks report performance gaps without probing underlying mechanisms. To address this, we introduce a benchmark with 1,657 questions across three foundational tasks designed specifically for mechanistic analysis. Examining model outputs across varying input settings (behavioral analysis) reveals that models often under-utilize audio when textual cues are available. We also provide the first causal mechanistic analysis of temporal reasoning failures in LALMs. Comparing attention upweighting against scaling, we find that redistributing attention across audio tokens is more effective than increasing audio attention. Targeting task-relevant tokens yields further gains. These findings suggest that modality imbalance alone cannot explain failures. Attention scaling at bottleneck layers improves accuracy from 55.9% to 59.1% without fine-tuning, demonstrating a promising direction for future work.

23.
arXiv (CS.LG) 2026-06-24

A Theory of Saddle Escape in Deep Nonlinear Networks

arXiv:2605.01288v3 Announce Type: replace Abstract: In deep networks with small initialization, training exhibits long plateaus separated by sharp feature-acquisition transitions. Whereas shallow nonlinear networks and deep linear networks are well studied, extending these analyses to deep nonlinear networks remains challenging. We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss and use this to classify activation functions into four universality classes. On the permutation-symmetric submanifold, the identity combines with an approximate balance law to reduce the full matrix flow to a scalar ODE, giving a critical-depth escape time law $\tau_\star = \Theta(\varepsilon^{-(r-2)})$ governed by the number $r$ of layers at the bottleneck scale rather than the total depth $L$. We find that this same $r-2$ exponent is recovered under He-normal initialization with $r$ bottleneck layers rescaled by $\varepsilon$, where the symmetry manifold is preserved by the flow but not attracting. We find close agreement between our theory and numerical simulations.

24.
PLOS Medicine 2026-06-18

Association between initial benzodiazepine prescribing patterns and time to benzodiazepine discontinuation: A population-based retrospective cohort study

by Nikki Bozinoff, Tanya S. Hauck, Robert A. Kleinman, Matthew E. Sloan, Beth A. Sproule, Simone N. Vigod, Jennifer Wyman, Priscila Pequeno, Tara Gomes Background Long-term benzodiazepine use has been associated with increased risk of morbidity and mortality. Preventing long-term use through safer prescribing practices has received little attention to date. We sought to better understand associations between initial prescription characteristics and duration of benzodiazepine use. Methods and findings This was a retrospective population-based cohort study of 1,820,808 adults in Ontario with incident benzodiazepine prescriptions between January 1, 2013 and December 31, 2020, with follow-up to December 31, 2021. The primary exposure was duration of the index prescription (≤7 days—referent group, 8–14 days, 15–30 days, or >30 days). Secondary exposures were: (a) duration of action of index benzodiazepine(s) prescription (short-acting, long-acting or both); (b) number of benzodiazepine dispensed on index (1 or 2+); and (c) mean daily dose of the index prescription in Diazepam Milligram Equivalents (DMEs). The primary outcome was time to benzodiazepine discontinuation in days. Multivariable models were adjusted for age, sex, anxiety, insomnia, and substance use disorders as well as other important comorbidities and socio-demographic characteristics. The median age at index was 53 years (Interquartile Range (IQR) 38–67), and 62.6% were women. The median time to discontinuation in women was 16 days (IQR: 6–29) while the median time to discontinuation in men was 19 days (IQR: 6–29). Lorazepam was the most commonly prescribed benzodiazepine on index (63.9%), followed by clonazepam (17.3%) and diazepam (5.8%). In multivariable Cox Proportional Hazards Models, longer index prescriptions were associated with a lower likelihood of benzodiazepine discontinuation (adjusted Hazard Ratio (aHR) 0.54 (95% Confidence Interval (CI) [0.54,0.54]) for 8–14 days; aHR 0.26 (95% CI [0.25,0.26] for 15–30 days and aHR 0.14 (95% CI [0.14,0.14]) for >30 days, compared to ≤7 days, respectively). Being prescribed two or more benzodiazepines versus 1 was also associated with a reduced likelihood of discontinuation (aHR 0.59 (95% CI [0.57,0.61])), as was being prescribed long-acting benzodiazepines (aHR 0.80 (95% CI [0.80,0.80])) or a combination of short and long acting benzodiazepine (aHR 0.84 (95% CI [0.80,0.88])) versus short-acting benzodiazepines alone. Mean daily doses of >5 to ≤10 DME and >10 to ≤20 DME were associated with an increased likelihood of discontinuation (aHR 1.03 (95% CI [1.03,1.03]); aHR: 1.03 (95% CI [1.03,1.04])), whereas doses >20 DME were associated with a reduced likelihood of discontinuation (aHR 0.98 (95% CI [0.97,0.98])) compared with ≤5 DME. Findings may be subject to bias from unmeasured confounding. Conclusion This large population-based cohort study found that prescribing shorter courses of benzodiazepines, use of a single benzodiazepine, use of a short-acting agent, were associated with reduced likelihood of long-term benzodiazepine use. Findings suggest that simple changes to prescribing practices could reduce prolonged benzodiazepine use and the morbidity and mortality associated with long-term use of these medications.

25.
arXiv (CS.CL) 2026-06-12

KCSAT-ML: Probing Reasoning Models with Nationwide-Cohort Human Difficulty

Math reasoning benchmarks have proliferated, yet most lack a per-item difficulty signal grounded in actual human performance. We introduce KCSAT-ML, a decade (2014-2025) of Korean College Scholastic Ability Test (KCSAT; Suneung) mathematics: 664 problems with a 339-item core set carrying official per-item error rates from nationwide cohorts of hundreds of thousands of examinees. We pair the benchmark with Difficulty-aligned Reasoning Gain (DRG): a score-orthogonal metric that asks whether a model's mistakes concentrate on the items humans found hard, or on items humans found easy. Together they expose, across a wide range of VLMs (and LLMs via OCR), three patterns: (i) low-budget accuracy collapses on the high-human-error tail at every model size; (ii) test-time scaling (TTS) raises token use roughly linearly with cohort error rate, while accuracy gains follow a non-monotonic curve; (iii) within a single family, TTS flips between anti-scaling on the hardest items and overthinking on easier ones – two faces of the same alignment failure. On DRG, models with near-identical accuracy can sit at near-opposite values: one model gets wrong what humans also find hard, while another solves the hardest items yet fails on items humans find easy – a contrast that aggregate accuracy hides. Our code and dataset builder will be open-sourced at https://github.com/naver-ai/KCSAT-ML.