Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
arXiv (CS.CL) 2026-06-11

Steering the Noise: Turning Random Perturbations into Effective Descent for Memory-Efficient LLM Fine-Tuning

Fine-tuning large language models (LLMs) achieves strong performance but is often limited by the memory overhead of backpropagation. Zeroth-order (ZO) optimization avoids this overhead by estimating gradients through forward passes alone, yet it typically converges slowly because random Gaussian perturbations yield high-variance gradient estimates in high-dimensional parameter spaces. In this paper, we propose a plug-and-play framework that turns random perturbations into more effective descent directions. The key idea is to draw a small pool of candidate perturbations, evaluate their loss values, and then select or combine those that are best aligned with the optimization objective. We develop two instantiations of this idea: MeZO-GV, which forms a guiding vector from the contrast between low-loss and high-loss perturbation groups, and MeZO-Greedy, which keeps the single best perturbation within a fixed evaluation budget. We theoretically show that both strategies yield a larger per-step reduction in the objective than standard ZO estimation, leading to improved convergence rates. Experiments on LLMs of different scales and architectures confirm that the proposed methods integrate naturally with existing ZO optimizers and consistently improve convergence speed and task accuracy. On OPT-13B, our approach outperforms all ZO baselines across 11 benchmarks and exceeds gradient-based methods on 9 of them, while retaining the memory efficiency of forward-only optimization.

02.
Nature (Science) 2026-06-22

Daily briefing: First-ever ‘nuclear’ clocks put atomic clocks in the shade

作者:

Two research teams have created a new, long-awaited type of timekeeper. Plus, how backlash has saved an ocean-monitoring network targeted by Trump and how our cultural heritage is put at risk by climate change. Two research teams have created a new, long-awaited type of timekeeper. Plus, how backlash has saved an ocean-monitoring network targeted by Trump and how our cultural heritage is put at risk by climate change.

03.
arXiv (CS.CV) 2026-06-11

Battery detection of XRay images using transfer learning

The need for detecting and sorting batteries is drastically increasing for many applications. This study proves the potential of transfer learning in predicting whether the image contains a battery or not, the location and identifying three types of batteries, namely: prismatic, pouch, and cylindrical Lithium-Ion Batteries (LIB). Particularly, it focuses on the transfer learning method in two applications: Training a large-scale dataset to detect electronic devices using a pre-trained YOLOv5m, then using these latter trained weights to detect and classify the batteries. The precision of battery detection achieves 94%, which outperforms the pretrained YOLOv5m weights with 5%, in 22 ms inference time.

04.
arXiv (CS.AI) 2026-06-19

On the Limitations of Ray-Tracing for Learning-Based RF Tasks in Urban Environments

arXiv:2507.19653v2 Announce Type: replace-cross Abstract: We study the realism of Sionna v1.0.2 ray-tracing for outdoor cellular links in central Rome. We use a real measurement set of 1,664 user-equipments (UEs) and six nominal base-station (BS) sites. Using these fixed positions we systematically vary the main simulation parameters, including path depth, diffuse/specular/refraction flags, carrier frequency, as well as antenna's properties like its altitude, radiation pattern, and orientation. Simulator fidelity is scored for each base station via Spearman correlation between measured and simulated powers, and by a fingerprint-based k-nearest-neighbor localization algorithm using RSSI-based fingerprints. Across all experiments, solver hyper-parameters are having immaterial effect on the chosen metrics. On the contrary, antenna locations and orientations prove decisive. By simple greedy optimization we improve the Spearman correlation by 5% to 130% for various base stations, while kNN-based localization error using only simulated data as reference points is decreased by one-third on real-world samples, while staying twice higher than the error with purely real data. Precise geometry and credible antenna models are therefore necessary but not sufficient; faithfully capturing the residual urban noise remains an open challenge for transferable, high-fidelity outdoor RF simulation.

05.
arXiv (CS.CV) 2026-06-18

Reasoning as Intersection: Consensus-Frame Alignment for Visual Focus in Video-MLLMs

Reinforcement learning has improved the reasoning ability of large language models, but applying outcome-only rewards to video multimodal large language models (Video-MLLMs) provides limited guidance on which visual evidence should support the answer. Inspired by multisensory integration, where consistent cues can enhance the salience and reliability of perceptual estimates, we introduce Consensus Frame GRPO (CF-GRPO), a temporal-annotation-free process-level reward framework for evidence-aware video reasoning. CF-GRPO constructs a consensus frame prior from intrinsic video cues, including temporal coverage, scene-transition cues, and query-conditioned visual relevance. It then computes a model-side frame-use score from visual and response representations and optimizes their agreement through the Consensus Frame Reward (CFR). With salience-aware sparse aggregation and distribution sharpening, CFR provides a high-contrast reward signal without requiring human temporal annotations. Experiments show that VideoCFR achieves competitive performance across complex video reasoning benchmarks and improves several metrics over representative Video-MLLM and RL baselines, while the consensus prior provides an interpretable view of the evidence frames emphasized during training. The implementation is available at https://github.com/1Pansy/VideoCFR.

06.
bioRxiv (Bioinfo) 2026-06-22

EMAlign: accurate alignment of cryo-EM maps through main-chain probability using deep learning

Accurate alignment of cryo-EM density maps is essential for comparing conformational states, searching map libraries, and guiding atomic model building, but remains challenging for noisy experimental maps and partially overlapping structures. Existing alignment methods are often based on raw maps, which may result in reduced accuracy due to the density noise, or require manual intervention for local alignment, which suffers from limited general applicability. Addressing the limitations, we present EMAlign, an automatic global and local cryo-EM map alignment with predicted main-chain probability using deep learning. First, EMAlign predicts main-chain prob ability maps from raw cryo-EM density maps using a BiMCUNet network. Then, a fast Fourier transform (FFT)-based search strategy is used to globally search the accurate alignment between cryo-EM maps based on predicted main-chain probability maps. As such, the main-chain prob ability map overcomes the noisy raw map problem, and the FFT-based exhaustive global search ensures the general applicability of alignment. EMAlign is evaluated on 64 global map pairs, 195 local map pairs, and 60 structure-to-map pairs at 3-10 [A] resolution and compared with gmfit, fitmap, VESPER, and CryoAlign. It is shown that EMAlign outperforms the other methods in both global and local alignment, achieving mean RMSDs of 1.03 [A] (global), 2.56 [A] (local), and 0.82 [A] (structure-to-map), with success rates of 100.0%, 100.0%, and 98.3% under the criterion of RMSD < 10 [A]. The EMAlign package is freely available at https://github.com/huang-laboratory/EMAlign/.

07.
arXiv (CS.CV) 2026-06-11

CountZES: Counting via Zero-Shot Exemplar Selection

Object counting in complex scenes is particularly challenging in the zero-shot (ZS) setting, where instances of unseen categories are counted using only a class name. Existing ZS counting methods that infer exemplars from text often rely on off-the-shelf open-vocabulary detectors (OVDs), which in dense scenes suffer from semantic noise, appearance variability, and multi-instance proposals. Alternatively, random image-patch sampling is employed, which fails to accurately delineate object instances. Since counting is sensitive to exemplar quality, such selection strategies often yield poorly representative exemplars, leading to inaccurate count estimation. To address these issues, we propose CountZES, an inference-only approach for object counting via ZS exemplar selection. CountZES discovers diverse exemplars through three synergistic stages: Detection-Anchored Exemplar (DAE), Density-Guided Exemplar (DGE), and Feature-Consensus Exemplar (FCE). DAE refines OVD detections to isolate precise single-instance exemplars. DGE introduces a density-driven, self-supervised paradigm to identify statistically consistent and semantically compact exemplars, while FCE reinforces visual coherence through feature-space clustering. Together, these stages yield a complementary exemplar set that balances textual grounding, count consistency, and feature representativeness. Experiments on diverse datasets demonstrate CountZES superior performance among ZOC methods while generalizing effectively across domains.

08.
arXiv (CS.CV) 2026-06-19

CMDS-AD: Cross-Modal Dual-Stream Decoupling for Few-Shot Anomaly Detection

Few-shot anomaly detection remains challenging due to limited training data. Multi-modal anomaly detection (MAD) offers a viable solution, leveraging 3D geometric cues to enrich 2D RGB representations and compensate for this scarcity. However, existing MAD methods apply spatially uniform feature processing, conflating stable macroscopic structures with high-frequency localized defect signals, exacerbating cross-modal misalignment and inflating false-positive rates. To overcome this, we present CMDS-AD, a Cross-Modal Dual-Stream Anomaly Detection framework. A LoRA-guided diffusion model generates diverse RGB samples to mitigate extreme data scarcity. For 3D normal augmentation, we employ a pre-trained diffusion model as a normal estimator. Crucially, this estimator inherently acts as a non-linear low-pass filter, directly extracting low-frequency normal representations from RGB inputs. This establishes an auxiliary estimated stream of purely low-frequency information, anchoring robust structural templates and assisting the uncompressed real stream, containing coupled high- and low-frequency components, to precisely isolate micro-defects. A Coordinate-Aware Hierarchical Feature Mapper adaptively aligns cross-modal semantics, while a multiplicative scoring mechanism filters modality-specific noise. Under the extreme 1-shot setting, CMDS-AD achieves absolute performance gains of 5.7% (I-AUROC) and 2.0% (AUPRO) on MVTec 3D-AD, alongside 7.7% and 5.6% improvements on EyeCandies, establishing a new state-of-the-art.

09.
arXiv (CS.CV) 2026-06-17

TerraTransfer: Learning End-to-End Driving Policies Without Expert Demonstrations

End-to-end autonomous driving has achieved state-of-the-art performance on benchmarks and real-world deployments. Its standard training recipe, however, is expensive across all stages: collecting and labeling millions of driving frames is costly, and closed-loop RL on images is bottlenecked by the per-step cost of photorealistic rendering plus a forward pass through a large vision backbone. Self-play in vectorized simulators changes the economics: millions of rollout steps per second, and a state distribution naturally rich in collisions, near-misses, and recoveries that no driving log contains. Our approach exploits this asymmetry by decoupling learning to drive from learning to see. We pretrain a single policy by self-play, then align its latent space with a pretrained vision backbone, through the action KL divergence and a batch-relational low-rank structural loss. The action target comes from the self-play policy, so alignment never supervises against a logged trajectory: a paired dataset of (image, scene-state) frames suffices, with no need for the curated expert demonstrations that imitation pretraining is built on. On photorealistic 3D Gaussian splatting closed-loop scenarios, the resulting end-to-end policy matches or exceeds prior end-to-end methods.

10.
Nature (Science) 2026-06-10

The Amazon can be saved — with concerted action inside and outside Brazil

作者: 未知作者

As deforestation in the Amazon falls, fresh evidence shows that the rainforest can withstand global warming, but only if there is a worldwide effort to stop cutting it down. As deforestation in the Amazon falls, fresh evidence shows that the rainforest can withstand global warming, but only if there is a worldwide effort to stop cutting it down.

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

MedicalAgentsBench for Complex Medical Reasoning: Comparing Internalized Reasoning Models versus Externalized Agent-based Frameworks

Complex medical reasoning requires integrating heterogeneous clinical evidence across multiple inference steps. Large language models (LLMs) now approach this through two routes: internalized reasoning and externalized agent scaffolding (frameworks that decompose problems collaboratively amongst multiple LLMs). To determine whether these routes are exclusive or complementary, we introduce MedicalAgentsBench, a filtered benchmark of 862 complex clinical questions drawn from the union of eight medical datasets via difficulty-aware curation and contamination screening. Evaluating three internalized reasoning models (DeepSeek-R1, o1-mini, and o3-mini), seven base models, and nine externalized agent-based methods, we find that internalized and externalized approaches each independently improve performance, and that their benefits compound: the highest accuracy is achieved by layering agent workflows onto an internalized reasoning model (i.e., o3-mini + MDAgents with 35.1%). Pareto analysis shows this combination dominates the cost-performance frontier; moreover, lightweight optimization on inexpensive models offers an entry point for resource-constrained settings. Our benchmark is at https://github.com/gersteinlab/MedicalAgentsBench.

12.
Nature Medicine 2026-06-15

Plasma proteomic signatures of cellular aging predict human disease

Aging is asynchronous across cells and organs. Here we tested whether plasma proteomics can be used to analyze cell type-specific aging. From analyses of over 7,000 plasma proteins measured in 60,542 individuals, we developed machine learning models to estimate the biological age of over 40 cell types spanning neuronal, immune, glial, endocrine, epithelial and musculoskeletal origins. We observed that 20–25% of individuals exhibited accelerated aging in a single cell type and 1–3% in 10 or more cell types. Cellular aging signatures were associated with disease status and predicted incident disease and mortality over 15 years of follow-up. Individuals with the APOE4 genotype showed older astrocytes but younger macrophages compared to APOE3 carriers, whereas the APOE2 genotype had inverse associations. Moreover, extreme astrocyte aging tripled the risk of incident Alzheimer’s Disease in individuals with two APOE4 alleles, while youthful astrocytes reduced risk. Individuals with extremely aged compared to youthful skeletal myocytes exhibited a 12.7-fold higher risk of developing amyotrophic lateral sclerosis. In individuals who smoked, extreme respiratory epithelial cell aging was associated with a 58% higher lung cancer risk compared to smoking alone. Specific cellular vulnerabilities and cumulative cellular aging burden influenced survival, with youthful immune and neuronal cell types conferring protective effects. Finally, we developed a polycellular aging risk score that stratified mortality risk across cohorts and proteomics platforms. These findings establish a framework for quantifying human physiology at cellular resolution, revealing heterogeneous aging trajectories and their impact on disease susceptibility and resilience. The biological age of individual cell types can be evaluated using plasma proteomics, revealing diverse aging profiles across more than 40 cell types and links between the accelerated aging of specific cell types and disease.

13.
arXiv (CS.LG) 2026-06-16

Integrated Marketing Attribution: A Bayesian Framework for Privacy-Safe Granular Measurement Anchored in MMM

arXiv:2606.16878v1 Announce Type: new Abstract: Retail marketing measurement increasingly requires granular campaign-level insights without relying on user-level tracking. However, the two dominant approaches, Marketing Mix Modeling (MMM) and Multi-Touch Attribution (MTA), often produce fragmented insights. MMM is privacy-safe and robust for channel-level planning but is too coarse for campaign optimization, while MTA provides granular attribution but has become less reliable under increasing privacy restrictions. We propose Integrated Marketing Attribution (IMA), a unified framework that combines MMM with channel specific Bayesian attribution models to derive campaign-level effects from aggregated data. By leveraging MMM-informed priors, IMA delivers granular, privacy-safe attribution while preserving consistency with MMM.

14.
bioRxiv (Bioinfo) 2026-06-22

Complex-valued representations of time-series gene expression profiles for network analysis

Time-series RNA sequencing provides a powerful framework for studying dynamic gene regulation, yet conventional analyses usually represent gene expression profiles as real-valued vectors in Euclidean space and quantify similarity using correlation or distance. Inspired by quantum information theory, we present a framework for encoding time-series gene expression profiles as complex-valued vectors comprising amplitude and phase components in Hilbert space. We designed multiple encoding models to represent gene expression in the amplitude of complex-valued vectors, encode temporal differences in the phase, and extend the phase representation to incorporate the direction of local expression changes. Gene-gene similarity was then quantified using fidelity, which measures the overlap between two encoded vectors. Evaluation using time-series RNA-seq datasets across diverse species and biological contexts showed that different encoding models produced distinct fidelity distributions that were related to, but distinct from, conventional correlation measures. We then constructed gene-gene networks using pairwise fidelity values and detected communities containing genes with similar temporal profiles. Although fidelity distributions differed across encoding models, the resulting communities captured major temporal expression programs, and functional annotations based on gene ontology and Kyoto encyclopedia of genes and genomes pathway analyses provided exploratory biological context. The detected communities were comparable to those obtained using conventional methods, including weighted correlation network analysis and fuzzy c-means clustering. Furthermore, as a proof-of-concept, we performed SWAP-test circuit simulations to mimic fidelity computation on a quantum computer; under noise-aware conditions, these simulations produced less accurate fidelity estimates with higher computational cost than classical computation. As a proof-of-concept, this study provides a complementary view of temporal transcriptome organization, rather than a uniformly superior alternative to conventional methods.

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

SING: Synthetic Intention Graph for Scalable Active Tool Discovery in LLM Agents

Large language model (LLM) agents increasingly rely on agent harnesses that manage context, tools, and multi-turn execution, making tools a central interface for acting in realistic digital environments. As harness-connected tool ecosystems expand to hundreds or thousands of APIs, services, and task-specific skills, exhaustive tool schema injection becomes costly and imposes a closed-world assumption that limits agents to a predefined static inventory. Retrieval-augmented tool selection offers a natural alternative, but existing one-shot retrieval methods often fail to align isolated tool descriptions with the agent's true task intention, especially in long-horizon tasks where required capabilities emerge through decomposition, observations, and newly induced subgoals. We propose SING, an intention-aware active tool discovery framework that builds an intention-tool graph linking user intentions, tool capabilities, and tool collaboration patterns, and dynamically retrieves tools according to evolving task states. Using a unified corpus of 7,471 tools, we evaluate SING on three real-world tool-use benchmarks. SING improves Global Recall@5 by up to 59.8% and downstream success rate by up to 28.9% over baselines, while reducing full-corpus tool-schema exposure by 99.8%, demonstrating that intention-aware graph structure enables more accurate and context-efficient tool discovery in large-scale agentic ecosystems.

16.
medRxiv (Medicine) 2026-06-24

Screen-Free Haptic Breathwork with HRV-Adaptive Control, Pilot Outcomes and System Design

Vayu is a mobile breathwork system comprising an iOS companion app and Apple Watch application that delivers slow, resonant breathing using screen-free haptic cues, HRV-adaptive pacing, and reflective journaling grounded in Patanjali's five states of mind. The watchOS component provides tactile phase guidance and real-time biometric sensing (heart rate, HRV), while the iOS interface supports analytics and personalized recommendations. In a 4-6-week naturalistic pilot involving 199 adults (ages 22-65) across Canada, the United States, and India, participants engaged in daily 5-10-minute sessions guided by on-wrist haptics. Average adherence was 4.1 +/- 2.3 sessions per week, with 71% of active users maintaining at least 3 sessions per week. By week four, perceived stress (PSS-10) decreased by 2.5 points, resting heart rate declined by 7.4 bpm, and HRV increased by a median of 28.6% relative to baseline, accompanied by mood improvements. No adverse events were reported. HRV metrics are derived from Apple Watch PPG-based proxies and interpreted as relative trends. These findings suggest Vayu is effective and well-tolerated, demonstrating strong engagement and early efficacy signals.

17.
arXiv (CS.CL) 2026-06-17

DICE: Diffusion Large Language Models Excel at Generating CUDA Kernels

Diffusion large language models (dLLMs) have emerged as a compelling alternative to autoregressive (AR) LLMs, owing to their capacity for parallel token generation. This paradigm is particularly well-suited for code generation, where holistic structural planning and non-sequential refinement are critical. Despite this potential, tailoring dLLMs for CUDA kernel generation remains challenging, obstructed not only by the high specialization but also by the severe lack of high-quality training data. To address these challenges, we construct CuKe, an augmented supervised fine-tuning dataset optimized for high-performance CUDA kernels. On top of it, we propose a bi-phase curated reinforcement learning (BiC-RL) framework consisting of a CUDA kernel infilling stage and an end-to-end CUDA kernel generation stage. Leveraging this training framework, we introduce DICE, a series of diffusion large language models designed for CUDA kernel generation, spanning three parameter scales, 1.7B, 4B, and 8B. Extensive experiments on KernelBench demonstrate that DICE significantly outperforms both autoregressive and diffusion LLMs of comparable scale, establishing a new state-of-the-art for CUDA kernel generation.

18.
arXiv (quant-ph) 2026-06-17

Response kinetic uncertainty relation for Markovian open quantum systems

arXiv:2501.04895v2 Announce Type: replace Abstract: Response uncertainty relations in stochastic thermodynamics extend precision bounds to the sensitivity of observables under external perturbations. Here we derive a quantum response kinetic uncertainty relation for continuously monitored Markovian open quantum systems in the steady state of the Lindblad master equation. The response precision of a measured trajectory observable is bounded by two contributions: the conventional quantum dynamical activity and a perturbation-induced intersubspace transition term. The latter is absent in the classical limit and captures a genuinely quantum part of the response cost. We identify simple conditions under which either contribution vanishes, and we further clarify the structure of the intersubspace term through a symmetry-resolved decomposition and exact sector-selection rules. The bound and its structure are illustrated in a driven two-level atom.

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

Robust Privacy: Inference-Stage Privacy through Certified Robustness

arXiv:2601.17360v2 Announce Type: replace-cross Abstract: An adversary observing a model's released prediction can infer sensitive attributes of the queried input, or even reconstruct representatives of the model's training data. The inference interface thus acts as a side channel for privacy leakage. We introduce Robust Privacy (RP), an inference-stage privacy notion inspired by certified robustness: if a model's prediction is provably invariant within a radius-R neighborhood around an input x with confidence at least $1-\alpha$, then x enjoys $(R,\alpha)$-Robust Privacy, under which we prove that any adversary observing the released prediction has at most $\alpha/2$ advantage in distinguishing x from any input within distance R of x. Building on RP, we formalize Robust Attribute Privacy (RAP), an attribute-level privacy notion that characterizes the set of sensitive-attribute values that remain compatible with a released prediction. On a classification task, RP increases the median length of the RAP-compatible inference interval from 23.50 to 29.96, reducing attribute-inference precision. Model inversion attacks, often treated as a training-stage threat, in fact rely on fine-grained signals leaked through the inference interface; RP masks these signals at the inference stage, reducing attack success rate (ASR) from 73% to 4% on a black-box inversion attack. This direct targeting of the leakage channel enables RP to dominate DP-SGD and randomized response in the privacy-utility tradeoff space: RP retains 98.4% accuracy at 21% ASR, whereas DP-SGD must drop accuracy to 61.7% to reach a comparable ASR. Across both experiments, increasing the smoothing sample size N strengthens privacy and improves utility together. Finally, we examine model distillation as a scope boundary and show that RP mitigates attribute-level and instance-level inference-stage privacy leakage, but not function-level extraction through model distillation.

20.
arXiv (CS.CV) 2026-06-18

Recognizing and Reconstructing a Multi-Unit Floor Plan

Digital twins have a major potential to form a significant part of urban management in emergency planning, as they allow more efficient designing of the escape routes, better orientation in exceptional situations, and faster rescue intervention. Nevertheless, creating the twins still remains a largely manual effort, due to a lack of 3D-representations, which are available only in limited amounts for some new buildings. Thus, in this paper we aim to synthesize 3D information from commonly available 2D architectural floor plans. We propose two novel pixel-wise segmentation methods based on the MDA-Unet and MACU-Net architectures with improved skip connections, an attention mechanism, and a training objective together with a reconstruction part of the pipeline, which vectorizes the segmented plans to create a 3D model. The proposed methods are compared with two other state-of-the-art techniques and several benchmark datasets. On the commonly used CubiCasa benchmark dataset, our methods have achieved the mean F1 score of 0.86 over five examined classes, outperforming the other pixel-wise approaches tested. We have also made our code publicly available to support research in the field.

21.
arXiv (CS.AI) 2026-06-16

Phys-JEPA: Physics-Informed Latent World Models for Multivariate Time-Series Forecasting

arXiv:2606.16076v1 Announce Type: cross Abstract: Multivariate forecasting in physical systems requires models that predict coupled temporal variables while preserving meaningful state evolution. Deep forecasters can fit temporal correlations, and physics-informed models can regularize predictions with scientific constraints, but these directions are often connected only at the decoded-output level. As a result, the hidden predictive state that generates future trajectories may remain statistically useful but physically unstructured. We introduce Phys-JEPA, a physics-informed joint-embedding predictive architecture for multivariate time-series forecasting. Phys-JEPA learns a latent world model in which predictive states are decomposed into physical and residual components, and physical consistency is imposed directly on latent states and latent transitions rather than only on decoded forecasts. This formulation uses known physical variables to organize the representation space while retaining residual capacity for unresolved dynamics. On Jena Climate 2009–2016, Phys-JEPA reduces aggregate MSE from 0.12482 to 0.12273 and temperature MSE from 0.01892 to 0.01831 at H=24. On Traffic, full Phys-JEPA improves aggregate MSE over the supervised baseline across all tested horizons, reducing H=192 MSE from 0.800784 to 0.773873. On Electricity, the best variant depends on horizon: static latent consistency is strongest at H=24 and H=48, while full Phys-JEPA gives the best aggregate and target-variable MSE at H=192. These initial results suggest that moving physics-informed learning from output space to latent predictive state space is a promising direction for interpretable temporal world models.

22.
arXiv (CS.LG) 2026-06-24

Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention

arXiv:2606.20945v2 Announce Type: replace Abstract: Self-attention is central to Transformer performance and is often the most expensive part of the Transformer at long context lengths because its pairwise token interactions scale quadratically with sequence length. Standard dense attention also applies the same set of attention heads to every token regardless of token difficulty or information content. This uniform activation can waste compute, especially as sequences grow longer and attention cost increases rapidly. We propose Grouped Query Experts (GQE), a mixture-of-experts layer on top of grouped-query attention (GQA). Within each GQA group, a router selects k query-head experts per token while all key-value (KV) heads remain dense and unchanged. Thus, GQE keeps the KV cache benefits of GQA and reduces only the active query-head computation. On a fixed 30B token budget at the 250M parameter scale, GQE matches the all-active GQA baseline in downstream accuracy while activating half the query heads per token.

23.
arXiv (CS.CV) 2026-06-19

Reliability-Aware Prototype Calibration for Frozen Pose-Flow Video Anomaly Detection

Pose-flow video anomaly detectors are attractive for one-class surveillance because they provide likelihood-based rankings for tracked skeleton windows. However, a single likelihood score may hide multimodal normal behavior and be sensitive to pose-observation noise. We study a frozen-detector setting in which the pose-flow backbone, cached skeleton tracks, and evaluation pipeline are fixed. Reliability-Aware Prototype Calibration (RPC) is a post-hoc score calibration method for this setting. It adds a standardized nearest-prototype deviation in the frozen latent space to the standardized flow score, and uses keypoint confidence only to gate this added geometric evidence. Thus, RPC preserves the original density signal while correcting the ranking with empirical normal-mode structure under pose reliability. Across two frozen pose-flow backbones and four datasets, RPC improves frame-level AUROC in all eight backbone-dataset pairs, with gains ranging from 0.34 to 4.49 percentage points and averaging 2.03 points. Ablation and reliability analyses show that prototype deviation is the main corrective signal, while reliability gating is most useful when pose observations are less trustworthy. These results suggest that lightweight post-hoc calibration can strengthen cached pose-flow systems when retraining or reproducing the full pose pipeline is impractical.

24.
arXiv (CS.AI) 2026-06-16

MUZZLE: Adaptive Agentic Red-Teaming of Web Agents Against Indirect Prompt Injection Attacks

arXiv:2602.09222v2 Announce Type: replace-cross Abstract: Large language model (LLM) based web agents are increasingly deployed to automate complex online tasks by directly interacting with web sites and performing actions on users' behalf. While these agents offer powerful capabilities, their design exposes them to indirect prompt injection attacks embedded in untrusted web content, enabling adversaries to hijack agent behavior and violate user intent. Despite growing awareness of this threat, existing evaluations rely on fixed attack templates, manually selected injection surfaces, or narrowly scoped scenarios, limiting their ability to capture realistic, adaptive attacks encountered in practice. We present MUZZLE, an automated agentic framework for evaluating the security of web agents against indirect prompt injection attacks. MUZZLE utilizes the agent's trajectories to automatically identify high-salience injection surfaces, and adaptively generate context-aware malicious instructions that target violations of confidentiality, integrity, and availability. Unlike prior approaches, MUZZLE adapts its attack strategy based on the agent's observed execution trajectory and iteratively refines attacks using feedback from failed executions. We evaluate MUZZLE across diverse web applications, user tasks, and agent configurations, demonstrating its ability to automatically and adaptively assess the security of web agents with minimal human intervention. Our results show that MUZZLE effectively discovers 44 new attacks on 4 web applications with 10 adversarial objectives that violate confidentiality, availability, or privacy properties across different LLMs and agent scaffolds. MUZZLE also identifies novel attack strategies, including 3 cross-application prompt injection attacks and an agent-tailored phishing scenario.

25.
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.