Sound
On Temporal Binding in Large Audio Language Models
Paul Primus and Gerhard Widmer
Reasoning about temporal structure of audio recordings requires Large Audio
Language Models (LALMs) to associate sound events with their temporal position. Understanding the underlying mechanisms is a first step toward diagnosing failures and identifying model components that may need improvement. Using mechanistic interpretability, we investigate how temporal information is represented and bound to sound events in three open-source LALMs. We find that across all three, event-specific location becomes concentrated in event name representations at intermediate modality integration layers. These representations encode coarse event position along a low-dimensional, curved relative time trajectory. Steering event name representations along this trajectory systematically shifts before/after beliefs, providing evidence that these representations contribute to coarse temporal
reasoning. In contrast, the same interventions do not reliably shift predicted onset timestamps, suggesting that coarse temporal
reasoning and precise metric event localization rely on distinct mechanisms.
Read the full abstract ▾
Artificial Intelligence
SIPO: Selective-Inference Policy Optimization for Tree-Structured Agentic RL
Zenghuang Fu, Ningqi Chen, Mingda Jia et al. (+10 more)
Tree-structured reinforcement learning trains search
agents by comparing alternative continuations and propagating terminal rewards to intermediate decisions. Adaptive expansion, however, creates a statistical asymmetry: an incumbent is selected using its own generation statistic, whereas fresh siblings are sampled after selection. When that statistic is associated with return, branch values can reflect selection history as well as continuation quality, even for a shared parent. We propose Selective-Inference Policy Optimization (\SIPO{}), which incorporates this distinction into tree-based credit estimation. Its scale-free branch criterion keeps generation scores and sibling penalties on a consistent relative scale; exchangeable branching supplies multiple fresh continuations from each selected parent; and order-statistic correction adjusts retained incumbent values using selection rank and the estimated score–outcome association. These mechanisms preserve the leaf budget and the host policy optimisation objective. Across seven QA
benchmarks using Qwen3-4B, Qwen3-8B, and Qwen2.5-7B, \SIPO{} achieves the highest reported multi-hop and single-hop averages among the compared methods. On Qwen3-8B, it improves these averages over AT\textsuperscript{2}PO by $1.31$ and $1.07$ percentage points, respectively, and ranks first on six of seven
benchmarks. Component ablations evaluate the individual and combined changes, while early-training paired diagnostics show a selected–fresh value gap alongside a near-zero fresh–fresh reference. Together, these results support accounting for selection history when constructing and evaluating search-
agent rollouts. Our code is available at https://github.com/Zenghuang-Fu/SIPO
Read the full abstract ▾
All authors Zenghuang Fu, Ningqi Chen, Mingda Jia, Xiaofeng Han, Zhaoyang Li, Qiuyuan Ai, Zelong Zheng, Haoyu Wu, Tianyu Fu, Chenxu Zhao, Minghui Wu, Guannan He, Changwei Wang
Cryptography and Security
Physics-Attested Federated Learning: Securing Collaborative Anomaly Detection in Critical Water Infrastructure
Jeff Nijsse, Shu Su, Benjamin Oholeguy et al. (+1 more)
Federated learning enables industrial operators to train shared intrusion detection models without disclosing proprietary operational telemetry. However, existing defences operate strictly in update space, leaving aggregators blind to data poisoning; model updates derived from fabricated telemetry remain indistinguishable from honest contributions. We repurpose process invariants, physical relations that always hold in normal operation, such as conservation of mass, from runtime detection heuristics into a verifiable admission requirement for
federated updates, mined automatically from clean operational data. We evaluate this admission gate across two physical water testbeds (SWaT, WADI) and a distribution
benchmark (BATADAL), testing seven aggregation rules against telemetry fabrication, exposure-only replay poisoning, and an invariant-aware adaptive adversary. Across three testbeds the mined invariants reject none of 25 honest shards except every cross-channel fabrication on SWaT and BATADAL, including optimised perturbations that the similarity based FoolsGold admits in full. On real telemetry, five mined invariants detect 12 of SWaT's 35 attacks, while nine invariants detect 20. With nine invariants, the admission gate removes 78% of the targeted-attack recall lost to replay poisoning (95% CI 57–97%, $p < 0.001$ over 20 seeds) across five standard aggregators, and 82% under FedAvg. To reconcile physical admission control with
federated data
privacy, we show invariant compliance using zero-knowledge proofs (zk-SNARKs) to allow clients to prove batch adherence without revealing operational telemetry.
Read the full abstract ▾
All authors Jeff Nijsse, Shu Su, Benjamin Oholeguy, Sreenivas Sremath Tirumala
Quantum Physics
General Photon Subtraction from a Gaussian Perspective
Niklas Budinger, Ulrik L. Andersen and Peter van Loock
A near-deterministic source of non-Gaussianity is the missing piece to reach universal and fault-tolerant
quantum computing with continuous-variable optics. Due to a lack of strong non-linearities, the probabilistic
photon-subtracted Gaussian states remain the most promising contender. In combination with Gaussian measurements and single-mode feed-forward operations, they have been shown to provide sufficient non-Gaussianity at high rates. However, these existing breeding protocols lack the necessary loss tolerance to be experimentally feasible. In this work, we introduce a mathematical formalism based on the application of a Gaussian blur and filter on the measured Fock Wigner function to describe general
photon-subtracted Gaussian states. Within our representation, the initial squeezing, Gaussian measurements, and
photon loss can all be understood as contributions to the total blurring and interchanged accordingly. We find that this intuitive approach can be used to find improvements to cat, cubic phase, and GKP state generation by compromising between success rates, feed-forward compatibility, and loss tolerance. In the case of multiple
photon subtractions, this is achieved by introducing a hybrid setup variant bridging the gap between protocols based on breeding and post-selection. Finally, we establish a general lower bound on the resources needed to generate a given target state near-deterministically by introducing the expected Wigner logarithmic negativity as an additive monotone of non-Gaussianity.
Read the full abstract ▾
Information Retrieval
No Attention, No Problem: Rethinking Session-based Recommendation with Pure Convolution
Tao Huang and Wei Zhou
Session-based recommendation (SBR) predicts the next choice in a session by analyzing recent interactions.
Transformer-based models are widely used because of their ability to capture long-range dependencies through self-
attention mechanisms. In contrast, traditional convolutional models, although more efficient, are often limited by their weak global modeling capabilities and are losing ground in SBR tasks. In this work, we propose a Next-generation Pure Convolutional Framework (NextConvRec) for SBR tasks, aiming to balance efficiency and performance. NextConvRec uses a Structural and Positional Convolutional Encoder (SPCE) for preprocessing, combining learnable convolutional positional biases with session-level structural signals extracted through GCN layers. Its backbone convolutional module effectively expands the effective receptive field through depthwise convolutions and pointwise convolutions, enabling robust long-range preference modeling without
attention mechanisms. Extensive experiments on 4
benchmark datasets show that NextConvRec outperforms several state-of-the-art baselines by around 1.73% on average, and reduces the average inference time per session by 16.7%. The convolutional architectures remain a promising direction for efficient and accurate session-based recommendations.
Read the full abstract ▾
Machine Learning
Do Personal Adapters Carry Over Across Sessions? Separating Personal From Population Gains When Calibrating EEG Foundation Models for Motor Imagery
Xilin Tao and Kani Chen
Every new user of a motor imagery
brain-computer interface must be calibrated, and users typically return on different days. A common approach with electroencephalography (
EEG) foundation models fits a small personal adapter on a frozen backbone, but evaluations are usually confined to one session and rarely separate the user's own contribution from that of the population model. We quantified personal adapters with two paired contrasts, the net gain of a user's own adapter over the population model and its advantage over adapters fitted to other users, for three frozen foundation models (CBraMod, REVE and LaBraM). In 235 held-out subjects from three
datasets, same-session calibration produced a positive net gain in all nine model and
dataset combinations (means of 1.5 to 5.4 percentage points), and the gain remained positive when population training was quadrupled. In 54 two-session subjects, adapters fitted on the first day still improved on the population model on the second day by 1.31 points for REVE and 1.64 points for LaBraM (Holm-adjusted p<0.01), retaining 61% and 83% of the gain from same-day recalibration, whereas the gain for CBraMod was 0.04 points and indistinguishable from zero. Same-day recalibration gave the highest mean accuracy for all three models, and adding 5 to 20 second-day labels to the first-day adapter mostly fell short of it. Whether personal adapters carry over across sessions therefore depends on the model and must be tested per model, and personalization should be reported against both a population reference and exchanged parameters, including across sessions.
Read the full abstract ▾
Computation and Language
Pass or Fail? Evaluating LLMs on Two Greek Examination Benchmarks
Panagiota Kyriazi, Eleni Kasoura and Prokopis Prokopidis
The rapid advancement of Large
Language Models (LLMs) imposes a thorough evaluation of their linguistic and analytical capabilities as well as constraints, particularly for a
language with limited
benchmark coverage such as Greek. To address the limited availability of comprehensive
benchmarks in this domain, we introduce Prot-Ex and Pan-Ex, two
benchmarks consisting of questions from entrance exams for Greek Model and Experimental schools as well as the Panhellenic exams (the Greek national university entrance examinations). These
benchmarks are employed to assess the performance of text-only LLMs-including the Greek-adapted KriKri-8B-Instruct, Llama-3.1-8B, Gemma-4-26B, and Qwen-3-32B-across diverse academic disciplines (Modern Greek, Mathematics, Physics, etc.) and task formats (closed, structured, and open-ended), including textualized visual context (i.e., image descriptions). Our findings indicate the localized KriKri-8B significantly outperforms its base model, successfully rivalling much larger LLMs in linguistically demanding humanities tasks. By leveraging an LLM-as-a-Judge methodology, we expose the inadequacy of traditional lexical metrics for evaluating complex
reasoning. Crucially, we uncover a few-shot prompting paradox: while synthetic examples improve accuracy in closed-ended questions, they severely overload the context window of 8B models in structured tasks, causing significant performance degradation. Ultimately, this study suggests targeted linguistic adaptation offsets lower parameter counts in specialized domains, despite the fragility of smaller models to
prompt verbosity.
Read the full abstract ▾
Artificial Intelligence
STRIDE: Automated Evaluation of Text-to-Trajectory Alignment across Diverse Contexts
Wanchun Ni, Tao Qi, Leonel Aguilar et al. (+4 more)
Language-conditioned trajectory generation is here, but its evaluation has not kept pace. Existing pedestrian trajectory metrics compare trajectories with real-world human data. This does not scale to text-to-trajectory generation across diverse contexts, as collecting human trajectories for every scenario is costly and infeasible. Moreover, pedestrian behavior is heterogeneous and context-dependent, with no single metric as the correct answer, and current evaluation frameworks are not transferable to this domain. These challenges make scalable, reliable evaluation difficult. We introduce STRIDE, the first framework for evaluating context
alignment between scenario descriptions and pedestrian trajectories. STRIDE addresses these challenges through three design choices. First, we derive our VRDST evaluation protocol from sociological theories to define a complete evaluation space. Second, it decomposes high-level context into scenario-adaptive behavioral questions. Third, every question is resolved against a deterministic measurement tool library that yields reproducible answers. Together, STRIDE enables verifiable, automated, and scalable evaluation across diverse contexts without requiring human trajectory data. We instantiate STRIDE in the crowd domain as STRIDE-Bench, comprising 1K scenarios, 6K behavioral questions, and 11K measurements with calibrated expected answers across 30 real-world maps. Comprehensive human validations show that STRIDE-Bench is consistent with human behavior and judgment, achieving 80% human agreement. We further evaluate several text-to-trajectory models, finding limited context-
alignment capability and persistent challenges in fine-grained context conditioning. We believe that the STRIDE framework provides a first step toward principled evaluation of context-aligned pedestrian trajectory generation.
Read the full abstract ▾
All authors Wanchun Ni, Tao Qi, Leonel Aguilar, Jiugeng Sun, Marlene Wagner, Verena Zimmermann, Mennatallah El-Assady
Computation and Language
InfiMed2: A Generalist Medical Multimodal Foundation Model from Contextual Evidence and Stability-Aware Supervision
Guanghao Zhu, Zeyu Liu, Zhitian Hou et al. (+9 more)
Recent medical
multimodal models have benefited from larger corpora, broader modality coverage, and stronger
reasoning-oriented training, yet effective data design across continued pretraining (CPT) and post-training remains challenging. Medical sources vary substantially in structure, granularity, and information density, and their utility shifts as training progresses from broad knowledge acquisition to late-stage consolidation. Meanwhile, post-training is often dominated by short-form visual question answering, providing limited supervision for informative and answer-consistent explanations. We introduce InfiMed2, a family of 4B and 27B generalist medical
multimodal foundation models built around stage-aware data design. We curate a 55.68B-
token corpus that combines broad clinical knowledge with context-rich biomedical visual evidence through source-specific processing. Our CPT pipeline first adapts the
vision encoder, then builds broad medical knowledge, and finally transitions to an evidence-focused data mixture during learning-rate decay. For supervised fine-tuning (SFT), we regenerate visual question-answering responses using answer stability, answer-masked reconstruction, and correctness-constrained selection to produce more informative and answer-consistent supervision. The 4B model is further optimized with reinforcement learning with verifiable rewards (RLVR). Across five medical
multimodal benchmarks, InfiMed2-4B achieves 66.73% mean accuracy after RLVR, surpassing the larger Qwen3.5-9B, while InfiMed2-27B reaches 73.72%, the highest among the evaluated open-weight models.
Read the full abstract ▾
All authors Guanghao Zhu, Zeyu Liu, Zhitian Hou, Pengkai Wang, Zhijie Sang, Shuo Cai, Yang Yu, Yuanyi Wang, Yanggan Gu, Congkai Xie, Jianmin Wu, Hongxia Yang
Signal Processing
Cross-Room Passive WiFi Tracking: Joint Estimation of Target Trajectory and Virtual Transmitter Location
Heping Wang, Zhongqin Wang, Henk Wymeersch et al. (+1 more)
Indoor bistatic WiFi sensing must often operate across rooms, where intervening walls block the direct line-of-sight (LOS) path between the transmitter (Tx) and receiver (Rx). Recovering a human trajectory from channel-state-information (CSI)-derived delay, angle-of-arrival (AoA), and Doppler estimates is difficult because the Tx position is often unknown and NLOS propagation is inconsistent with direct-path geometry. This paper presents CoTrack, a self-calibrating passive WiFi tracking scheme that does not require the Tx position. CoTrack extracts one dominant human-induced response and represents its transmitter-side propagation by a virtual Tx, which it estimates jointly with the human trajectory. Because similar measurements can be explained by different virtual-Tx–trajectory pairs, CoTrack uses multi-start nonlinear least squares and accepts an initialization only when the best-fitting candidates concentrate around a common virtual-Tx position. It then applies online alternating optimization for continuous tracking. We analyze the local identifiability of the joint estimation problem and derive the corresponding Cramér–Rao bound. Across six LOS and cross-room NLOS experiments, CoTrack achieves a median trajectory error of 1.14m(80th percentile: 1.42m) and a median virtual-Tx discrepancy of 0.47m(80th percentile: 0.54m).
Read the full abstract ▾
All authors Heping Wang, Zhongqin Wang, Henk Wymeersch, J. Andrew Zhang
Materials Science
Phase compositions of Si oxynitride: A comprehensive thermodynamic modeling study
Andrey Sarikov, Mariia Babiichuk, Mariia Voitovych et al. (+1 more)
In this work, a comprehensive thermodynamic description of the formation of phases in amorphous SiOxNy (0 <= x <= 2, 0 <= y <= 4/3) films as a function of the film chemical composition and fabrication temperature is presented. Two possible phase states, namely homogeneous single-phase Si oxynitride and two-phase mixture of Si
oxide and Si nitride phases, are analyzed in terms of the lowest values of the Gibbs free energy. The characters of the modification of the phase composition upon changing the SiOxNy stoichiometry indices and fabrication temperature are determined. Moreover, characteristic features of the chemical composition of the Si
oxide and Si nitride phases in two-phase Si oxynitrides are revealed. Particular
attention is given to the thermodynamic origin of the boundaries between different phase states of the SiOxNy films and to determining the critical stoichiometry values separating them. A comprehensive phase diagram indicating single- and two-phase SiOxNy regions as functions of the fabrication temperature is constructed. The peculiarities of the transitions between different SiOxNy phase states, along with underlying mechanisms, are discussed with reference to this phase diagram. The obtained results extend our understanding of the fundamental properties of Si oxynitride films and their dependence on the film fabrication conditions, and may be used for tailoring SiOxNy characteristics for practical applications.
Read the full abstract ▾
All authors Andrey Sarikov, Mariia Babiichuk, Mariia Voitovych, Sergii V. Mamykin1
Computer Vision
Physics-Guided Spectral Distillation for Underwater Image Enhancement on Resource-Constrained Devices
Yifan Chen, Kai He, Ye Zheng et al. (+3 more)
Underwater image enhancement is crucial for improving visual perception in marine applications. Existing
underwater image enhancement studies mainly focus on enhancement quality and visual fidelity, while rarely considering real-time deployment capability, which is essential for resource-constrained
underwater robots. To this end, we introduce a physics-guided spectral
distillation (PSD) method, which reduces model capacity for real-time applications while maintaining the high performance of
underwater image enhancement models. To decompose the outputs of teacher and student models, PSD adopts a multilevel Haar discrete wavelet transform. It transfers low-frequency color and illumination information as well as high-frequency structural details through band-specific objectives. Moreover, the
distillation process of PSD is degradation-aware. We estimate degradation-aware weights through a physical head and combine them with ground-truth-guided reliability masks to selectively retain valuable teacher guidance. Experiments on the UIEB, LSUI, and EUVP
datasets validate the effectiveness of the proposed method. Furthermore, we demonstrate the benefits of enhanced images for downstream perception tasks, including object detection. Deployment on a self-developed ROV further demonstrates its practical applicability in real-world
underwater scenarios.
Read the full abstract ▾
All authors Yifan Chen, Kai He, Ye Zheng, Jijun Lu, Zhe Sun, Tao Chen
Computational Engineering
Where Do Embodied Decisions Come From? Rethinking Latent and Explicit Reasoning
Yuan Lin, Ziyue Zhou, JinLong Zhao et al. (+4 more)
Chain-of-Thought (CoT)
reasoning is increasingly incorporated into
Vision-
Language-Action (VLA) models, yet it can degrade the performance of stronger
embodied agents. We investigate this capability-dependent effect by distinguishing explicit
reasoning from latent decision computation, i.e., perception-grounded computation that directly supports action prediction. Under the standard think-then-act (TTA) paradigm, intervening on the generated CoT while fixing the visual input and model parameters causes a substantial performance collapse, with
Navigation F1 dropping from 72.14% to 11.84%, demonstrating the strong influence of explicit
reasoning on action generation. We then propose decide-then-explain (DTE), which predicts actions before generating explanations, and introduce Visual Conditional Contribution (VCC) and
Reasoning Conditional Contribution (RCC) to characterize the resulting decision process. Across autonomous driving and robotic manipulation, DTE consistently outperforms TTA and conventional no-CoT baselines, while exhibiting greater reliance on perception-grounded computation. Further TTA-trained, DTE-inference experiments show that this benefit is not solely attributable to retraining under the new factorization. Our results suggest that for capable
embodied agents, explicit CoT may be better used to shape decision computation during training rather than mediate action generation at inference time. Code: https://github.com/ocean-luna/openvla-decide-then-explain.
Read the full abstract ▾
All authors Yuan Lin, Ziyue Zhou, JinLong Zhao, Pei Liu, Haipeng Liu, Pan Zhou, Kun Zhan
Quantum Physics
Efficient Simulation of Hybrid Continuous- and Discrete-Variable Quantum Circuits via Gaussian Decompositions
Kazufumi Tanji, Dot Belin Pio, Shohei Kiryu et al. (+2 more)
Hybrid
quantum systems combining discrete-variable (DV) and continuous-variable (CV) subsystems arise across a broad range of physical platforms and enable
quantum information processing beyond purely
qubit-based designs. Their classical simulation, however, is challenging since computational costs of both CV and DV systems exponentially increase. In this paper, we introduce a simulation framework for CV–DV hybrid systems based on linear combination of Gaussian (LCoG) decomposition. We decompose the hybrid density
matrix into blocks in a DV basis and represent each block as an LCoG expansion. Within the representation, Gaussian operations on the bosonic subsystem independently act on each Gaussian function. We derive analytic formulas describing how complex Gaussian functions transform under controlled Gaussian operations. We further extend the framework to controlled Gaussian channels with state-dependent Gaussian noise. The method avoids a Fock-space truncation for the CV subsystem, and the cost of updating each Gaussian term scales polynomially with the number of modes. The framework provides a general computational tool for analyzing hybrid
quantum algorithms,
quantum simulation protocols, and non-Gaussian state generation schemes. As an example, we simulate Gottesman–Kitaev–Preskill (GKP) state generation based on a cavity-QED system.
Read the full abstract ▾
All authors Kazufumi Tanji, Dot Belin Pio, Shohei Kiryu, Ulrik Lund Andersen, Masahiro Takeoka
Computer Vision
D$^2$-VLA: Dual-Memory Dual-Frequency Vision-Language-Action Model For Long Dynamic Manipulation
Zijian Ye, Chengqi Wei, Wei Huang et al. (+9 more)
Long-horizon manipulation requires
robots to remember cues that are no longer in view while responding to moving objects. Yet
vision-
language-action (VLA) policies often rely on the latest observation, and refreshing their visual context typically requires another costly
vision-
language model (VLM) pass. We present D$^2$-VLA, which combines dual
memory and dual-frequency control at the KV-cache interface of a pretrained VLA. D$^2$-VLA uses block-wise causal KV caching to encode observations incrementally and, guided by distinct temporal
attention patterns, constructs separate historical KV read views for the VLM and action expert. Between periodic VLM updates, a gated adapter incorporates fresh visual features into the latest history-conditioned KV block, while a short fast-
memory queue supports action replanning. We introduce DOMINO-Long, a ten-task
benchmark requiring
robots to use earlier visual cues when manipulating moving objects. D$^2$-VLA achieves complete-task success rates of 29.3% on DOMINO, compared with 9.6% for $\pi_{0.5}$ and 17.2% for PUMA, and 60.0% on DOMINO-Long, compared with 35.4% and 20.6%, respectively. It improves success rates on eight real-
robot tasks and reaches 97.5% on LIBERO-Long and 74.3% on RoboTwin 2.0.
Read the full abstract ▾
All authors Zijian Ye, Chengqi Wei, Wei Huang, Anlin Zheng, Chunyu Zou, Liangyu Wu, Zikang Zhao, Zhenjie Peng, Yushuo Yang, Shuman Zhao, Zhongrui Wang, Xiaojuan Qi
Optimization and Control
Finite-Time Concentration and Convergence Rates for Projected Two-Time-Scale Stochastic Approximation with Markov Noise
Rahul Singh, Vivek S. Borkar and Eric Moulines
We study finite-time concentration and convergence rates for projected two-time-scale
stochastic approximation driven by a controlled Markov chain. The averaged fast map is contractive, while the slow iterate is projected onto a compact convex polyhedron. The associated projected ordinary differential equation may have a discontinuous vector field at the boundary, preventing a direct application of standard analyses based on Lipschitz vector fields. Using the Skorokhod map, we establish explicit high-probability bounds for tracking the moving fast
equilibrium and the projected slow dynamics. These bounds separate martingale fluctuations, Markov-noise residuals, and the bias due to time-scale separation. A Lipschitz Lyapunov function satisfying a uniform decrease condition over fixed time intervals yields almost-sure convergence, with explicit last-iterate rates when the decrease admits a power lower bound. Under uniform Lyapunov contraction,
polynomial step sizes yield joint fast-tracking and slow Lyapunov-error exponents arbitrarily close to $1/3$. Under the additional assumption that the reduced slow update map is a Euclidean contraction, logarithmically separated step sizes improve the joint rate to $O(n^{-1/2}\log n)$ almost surely, including for boundary
equilibria. The same rate holds under a distinct
geometric condition involving a strictly attracting face of a box and a fast
equilibrium that is constant on that face. An actor-critic application achieves an almost-sure value-gap rate of $O(n^{-1}\log n)$ relative to the optimum within the constrained policy class. Further applications include projected TD(0) and projected
stochastic gradient descent. We also extend the analysis to projection of the fast recursion under Euclidean contractivity.
Read the full abstract ▾
Cryptography and Security
CoSec: Benchmarking Agent Security in Communities
Hao Chen, Wenhui Dong, Ye Chen et al. (+12 more)
LLM
agents operate in persistent collaborative environments involving multiple users, communities,
memories, files, and tools. Community boundaries may remain fixed or evolve with changes in membership, roles, composition, and relationships.
Agents must complete legitimate tasks and prevent unauthorized disclosure of protected information. Existing evaluations do not fully examine these risks in
agent systems. We introduce CoSec, an executable
benchmark for evaluating
privacy and authorization enforcement in LLM
agent systems operating within and across communities. CoSec contains 208 canonical scenarios spanning fixed and evolving boundaries, protected information belonging to the
agent owner or other participants, and attacks through dialogue, environmental content, persistent
memory, and composed workflows. CoSec executes complete
agent systems with persistent sessions,
memory, files and tools. It verifies information flows against the active authorization state using execution traces and artifacts. Across harness and model configurations,
agents frequently complete benign tasks but violate
privacy and authorization boundaries.
Privacy behavior varies across harnesses, attack surfaces, and community states, revealing how
memory, files, tools, and workflows can carry protected information beyond its authorized scope. These findings show that task utility does not imply
privacy or authorization compliance and that authorization in community settings remains an unresolved security challenge for persistent LLM
agents.
Read the full abstract ▾
All authors Hao Chen, Wenhui Dong, Ye Chen, Jiezhi Yao, Chenbo Xia, Yuwen Qu, Renxiang Wang, Fudong Yuan, Camil Hamami, Chenglong Pan, Xinquan Yue, Ziyu Wang, Fengyu Ye, Chenyang Si, Caifeng Shan
Databases
Transformations for Evolving Property Graph Schemas
Gauvain Devillez, Stefania Dumbrava and Angela Bonifati
Property graph databases are widely used to represent complex and evolving data, yet systematic support for property graph schema
evolution remains limited. In practice, schema transformations are typically defined manually, coupled to specific application contexts, and difficult to reuse across schemas or
evolution scenarios. We present GRAFT, a logic-based framework that models property graph schema
evolution as reusable, order-constrained meta-transformations derived from atomic edits. Schema
evolution is formulated as the exploration of a finite meta-graph with schemas as nodes and grounded meta-transformations as edges. To ensure tractability, GRAFT combines similarity-guided search and
pruning, guaranteeing duplication-freeness, termination, and correctness. An experimental evaluation on four
benchmark and real-world property graph schema
evolution scenarios shows that GRAFT efficiently computes high-quality schema transformation sequences. Using greedy exploration, GRAFT reaches the exact target schema on most
datasets, producing stable transformation sequences while keeping runtimes low. A qualitative study on both real-world and synthetic large-scale
datasets further shows the quality and robustness of the obtained reusable meta-transformations.
Read the full abstract ▾
Cosmology and Nongalactic Astrophysics
The separate universe approach with non minimal coupling to gravity
Hugo Holland and Julien Grain
In this paper we extend a study of the validity conditions of the separate universe approach of cosmological perturbations to models of inflation with multiple fields that are non minimally coupled to
gravity. The separate universe approach effectively describes the universe as a collection of homogeneous and isotropic patches, leading to an effective description of cosmological perturbations at large scales. This approximation is a necessary step in
stochastic inflation and in the $\delta \N$ formalism. We study and compare the dynamics of cosmological perturbation theory at large scales to the dynamics derived in the separate universe picture for non linear sigma models where we have allowed a non minimal coupling between the fields and
gravity. We found that the separate universe approach adequately reproduces the dynamics of the large scale perturbations, provided the wavelength of the modes considered is greater than a list of lower bounds derived from the background
evolution of the model we consider. These lower bounds are then compared to those derived in the classically equivalent Einstein frame description of the Jordan frame model at hand. In particular these conditions are derived and compared for a
Higgs-like toy model of inflation in the large field limit.
Read the full abstract ▾
Differential Geometry
Rigidity of Lagrangian submanifolds with conformal Maslov form and Legendrian capillary boundary
Dong Gao, Yong Luo, Hui Ma et al. (+1 more)
We classify smoothly immersed Lagrangian $n$-balls, $n\geq 2$, in the unit ball of $\mathbb{C}^n$ with conformal Maslov form and Legendrian capillary boundary. The image of every such immersion is either an equatorial Lagrangian disk or is contained in a Whitney sphere centered at the origin. In dimension two, this confirms a
conjecture of Li, Wang and Weng [Sci. China Math. 2021].
Read the full abstract ▾
All authors Dong Gao, Yong Luo, Hui Ma, Jiabin Yin
Machine Learning
Instance-Adaptive Prompts as Context for Time-Series Foundation Models
Zehao Xiao, Shifeng Xie, Lei Zan et al. (+5 more)
Longer histories can improve time-series foundation models (TSFMs), but require substantially higher inference cost. We therefore ask whether contextual information can be provided more efficiently through a compact set of learned
token embeddings. We introduce PaCTS, which generates a small set of instance-adaptive latent
prompts in the form of continuous
embedding tokens conditioned on the visible context. These
prompts serve as compact context surrogates for frozen TSFMs. PaCTS constructs them from instance-specific global statistics and further refines them with segment-level temporal information, capturing both global characteristics and local temporal variations. The
prompt module is jointly trained and deployed across heterogeneous time series with the frozen backbone. Extensive experiments demonstrate the effectiveness of
prompts as context, consistently improving
forecasting across context lengths and model architectures. With a shorter input context, PaCTS can outperform the same frozen backbone using double context while requiring substantially less inference computation. Compared with weight-space adaptation methods, PaCTS achieves stronger improvements and better out-of-distribution generalization.
Read the full abstract ▾
All authors Zehao Xiao, Shifeng Xie, Lei Zan, Jianfeng Zhang, Lujia Pan, Ievgen Redko, Malik Tiomoko, Keli Zhang
Artificial Intelligence
BEHAVE: Functional Behavior Modeling Enables Self-Improving Agents for Hardware Design and Verification
Yuheng Wu, Berk Gokmen, Sujeeth Jinesh et al. (+5 more)
Developing
agents for hardware design and verification requires reliable correctness feedback. As a hardware specification may permit correct implementations with different latencies, matching design and reference outputs cycle by cycle can reject valid designs. To address this, we introduce BEHAVE, an agentic framework for multi-turn joint hardware design and verification through functional behavior modeling. We define Behavior IR to express task functionality as executable behavior models without prescribing implementation timing beyond the specification. The
agent iteratively develops a register-transfer-level (RTL) design and a behavior model as the design's verification reference. Our evaluator, BEHAVE-Sim, checks both artifacts separately against a hidden golden behavior model using input stimuli generated by random sampling and solver-guided search. BEHAVE thus supports power, performance, and area (PPA) exploration across task-permitted latencies and microarchitectures. During training, the same evaluator provides verifiable reinforcement learning (RL) rewards from specification-behavior pairs without reference RTL. For self-improvement, the
agent continually searches for high-level implementations relevant to its capability gaps, constructs and checks specification-behavior pairs, and trains on the expanded task pool. We release BEHAVE-Train and BEHAVE-Eval with 600 human-reviewed specification-behavior pairs for realistic hardware workloads. Starting from 60 seed tasks and acquiring 100 new tasks, self-improvement raises Qwen3.8-27B's RTL pass@1 on BEHAVE-Eval from 55.0% to 75.0%, reaching performance comparable to RL using a 540-task pool.
Read the full abstract ▾
All authors Yuheng Wu, Berk Gokmen, Sujeeth Jinesh, Lauren McLane, Aarav Wattal, Qi Yang Huang, Zhaozhuo Xu, Thierry Tambe
Computer Vision
Projective Normal Fields: A Convex Optimization Method for Constructing Smooth UDFs
Jiayi Kong, Chen Zong, Fei Hou et al. (+3 more)
Constructing a smooth approximation of an unsigned distance field (UDF) from a raw point cloud is challenging because the input provides neither surface connectivity nor consistently oriented normals. Methods that directly learn a scalar UDF must also handle its non-differentiability on the zero level set and weak supervision away from the samples, which can lead to unstable optimization and spatial artifacts. We introduce Projective Normal Fields (PNFs), an orientation-free representation and convex optimization framework for estimating bidirectional normals from point positions alone. Each normal axis is encoded by a rank-one projector, which is invariant to normal reversal. We relax the non-convex set of hard projectors to its convex hull: the symmetric positive-semidefinite
matrices with unit trace. Each soft tensor defines a local quadratic distance model and retains the relative weights of candidate normal axes. We estimate a coherent PNF by combining local tangent-plane fitting, soft-PCA anchoring, and overlap regularization on a fixed neighborhood graph. With positive anchoring weights, the objective is strongly convex and admits a unique global minimizer. Principal eigenvectors provide bidirectional normals, while the corresponding eigengaps provide spectral confidence indicators. We use these indicators to select and weight directional sources for heat
diffusion, followed by Poisson integration to construct a regularized UDF approximation. By separating local geometry estimation from scalar-field construction, PNF avoids directly fitting the non-differentiable UDF. Experiments demonstrate reduced sensitivity to neighborhood size, competitive reconstruction under noise and outliers, and improved accuracy near non-
manifold junctions. The project page is available at https://anonymous17777367.github.io/PNF-page/
Read the full abstract ▾
All authors Jiayi Kong, Chen Zong, Fei Hou, Junhui Hou, Wenping Wang, Ying He
Computation and Language
TQTS-Bench: A Multi-Syntax Benchmark for Text-to-Query over Time-Series Databases
Fei Lyu, Zhiyi Peng, Jiaming Liu et al. (+5 more)
Large
language models (LLMs) have significantly advanced natural
language querying over relational databases, yet their ability to query time-series databases (TSDBs) remains largely unassessed. Existing
benchmarks fail to adequately capture the non-unified query syntaxes, diverse application domains, and unique time-specific query intents inherent to TSDBs. To address this gap, we introduce TQTS-BENCH, a multi-syntax
benchmark for evaluating text-to-query capabilities over TSDBs. TQTS-BENCH contains 6,125 high-quality question-answering (QA) pairs spanning 97 TSDBs, 23 distinct query syntaxes, 22 application domains, and 4 types of time-specific query intents. It is constructed through a human-centric AI-assisted workflow, where all QA pairs are carefully reviewed and revised by domain experts to ensure quality and correctness. Extensive evaluations of advanced LLMs and state-of-the-art text-to-query methods reveal challenges in querying TSDBs. Even the best-performing model evaluated, Claude-Opus-5, achieves only 48.98% execution accuracy, while humans reach 87.34%. Error analysis reveals that this performance gap mainly stems from the heterogeneous query syntaxes across different TSDBs, misinterpretation of time-specific intents, and incorrect schema linking. These findings highlight new opportunities to narrow the gap between current LLM capabilities and the requirements of TSDB queries in real-world applications. The
benchmark is available at: https://anonymous.4open.science/r/TQTS-Bench-00CD.
Read the full abstract ▾
All authors Fei Lyu, Zhiyi Peng, Jiaming Liu, Yixuan Yang, Changjian Chen, Zhuo Tang, Jiapeng Zhang, Kenli Li
Robotics
CoHuB: A Simulation Benchmark for Multi-Humanoid Collaboration
Hyunjin Park, Jebeom Chae, Minwoo Park et al. (+11 more)
Many physical tasks in human environments require collaboration, from assisting a partner to jointly manipulating an object. Yet, existing
humanoid benchmarks largely focus on single-
humanoid skills and lack evaluation of multi-
humanoid collaboration under
egocentric visual observations. We introduce CoHuB (Collaborative Multi-
Humanoid Benchmark), a simulation
benchmark for multi-
humanoid collaboration under
egocentric visual observations. CoHuB provides 10 tasks, eight with two
humanoids and two with three
humanoids, spanning diverse collaboration patterns. We also provide synchronized demonstrations collected through a multi-operator VR teleoperation pipeline, in which each operator controls one
humanoid from its
egocentric view. Experiments with representative visuomotor policies reveal substantial challenges across different forms of coordinated perception and control. CoHuB provides a foundation for developing and evaluating multi-
humanoid collaboration policies.
Read the full abstract ▾
All authors Hyunjin Park, Jebeom Chae, Minwoo Park, Sunghyun Park, Hanjun Yoo, Seoyeon Choi, Soochul Yoo, Joohwan Seo, Sarmad Idrees, Jae-Sang Hyun, Jongmin Lee, Roberto Horowitz, Youngwoon Lee, Jongeun Choi
Computer Vision
When VLMs Trust Context: Evaluating Scene Text Recognition under Misleading Context
Yuxing Cheng, Yuan Wu and Yi Chang
Vision-
language models (VLMs) can read text in natural scenes, but their predictions may be influenced by the surrounding context. When the printed text conflicts with what the scene suggests, a model may return a more plausible word instead of the shown text. We introduce SceneFaith, a
benchmark of 781 generated scene images for studying this behavior. Each output is classified as Literal, Canonical, or Other, separating faithful transcription from context-consistent rewriting and ordinary recognition errors. Across 15 models from seven families, all models show rewriting on clear images, with rates ranging from 8.45% to 58.51%. Controlled experiments further show that surrounding context matters: removing surrounding scene information reduces rewriting and improves literal accuracy, while changing the scene around the same text patch can also change model outputs. Moreover, weakening the target text with blur increases rewriting. These results show that reliable scene-text recognition requires VLMs to balance visual character evidence with contextual information, preserving clear text while using context mainly when the visual evidence is uncertain.
Read the full abstract ▾
Artificial Intelligence
Applying Language Models in Clinical Medicine: Recent Trends and Perspectives
Erik Aerts
The use and applicability of artificial intelligence (AI) in medical research and clinical practice has received increasing
attention in the literature over recent years. The emergence of large
language models (LLMs) has expanded discussions in regards to applications of AI within healthcare. While traditional deep learning based AI applications in medicine have often focused on specific and defined tasks, LLMs offer broader capabilities and flexibility in working with available data,. At the same time of writing, the integration of LLMs into medical settings raises important questions regarding their reliability, accuracy, transparency, safety, and appropriate role in a medical setting. This text presents and discusses recent talks and articles concerning the application of LLMs in medicine, with particular emphasis on their potential utility in research and clinical practice. It considers both the opportunities offered by these technologies and the challenges associated with their implementation, aiming to provide a perspective on the current and emerging role of LLMs within the medical field.
Read the full abstract ▾
Information Theory
Divergence-Minimizing Distribution Matching for Parallel Channels and Channels with Memory
Francesca Diedolo and Gerhard Kramer
Theory for distribution matchers (DMs) is extended to distributions with
memory. The divergence-minimizing DM is shown to select the sequences with the highest target probability, as in the memoryless case. A scaling law for divergence is extended to distributions driven by innovation processes. The theory is applied to parallel additive white Gaussian noise channels. A modified enumerative sphere-shaping (ESS) method with a weighted energy constraint is used in implementations. An illustrative example with three channels shows that joint ESS across channels reduces the rate loss by a large factor compared to product DMs at short blocklengths. The gains are confirmed by simulations with probabilistic amplitude shaping and a 5G-NR low-density parity-check code.
Read the full abstract ▾
Software Engineering
TraceLib: System-Call Bitmap Feedback Mechanism for Language-Agnostic Web Fuzzing
I Putu Arya Dharmaadi, Elias Athanasopoulos and Fatih Turkmen
Coverage feedback is an important source of guidance for fuzzing. However, obtaining such feedback normally requires application-level instrumentation that is specific to the
language and runtime of the application. Given that modern web applications span multiple
languages and runtimes, this application-level instrumentation is costly to implement and maintain. Therefore, we present TraceLib, a system-call feedback mechanism for enabling
language-agnostic web fuzzing. Our proposed approach observes the transitions of system calls, enriches selected transitions with bounded argument hashes, and converts them into a 65,536-position AFL-like bitmap. By using the generated bitmap, any web fuzzer can decide whether to retain requests that add previously unseen bitmap positions without consulting application code coverage. We integrate TraceLib into WebFuzz and evaluate it under five WebFuzz feedback modes: the two proposed TraceLib variants (one over every traced system call and one projected onto monitored file paths and recognized SQL buffers), the N-gram comparator adapted from Xiao et al. representing the most recent work to our knowledge, WebFuzz's Native grey-box feedback, and black-box fuzzing without feedback. We evaluate TraceLib on sixteen web applications under test (WUTs): eight PHP applications and eight further applications spanning Node.js, Ruby, Java, Go, and Python to demonstrate platform portability. The results show that our proposed TraceLib projected exceeds black-box on all eight PHP WUTs while exceeding Native on four: Joomla, Drupal, PrestaShop, and Bagisto. In addition, measured on an identical replayed request workload, the tracer costs approximately one millisecond of server-side latency per request. These results indicate that compact system-call feedback is a useful runtime-independent proxy for coverage guidance.
Read the full abstract ▾
High Energy Physics (Theory)
M2 brane $S^3/\mathbb Z_k$ instanton partition function in ${\rm AdS}_4\times S^7/\mathbb Z_k$: 2-loop correction
Arkady A. Tseytlin
As was shown in arXiv:2609.14497, the 2-loop $ {\rm T}^{-1}_2$ correction to the partition function of M2 brane wrapping $\mathrm{AdS}_2\times S^1$ inside $\mathrm{AdS}_4\times S^7/\mathbb{Z}_k$ vanishes. This is in agreement with the localization prediction for $\frac{1}{2}$-BPS circular Wilson loop in ABJM theory interpreted in the grand-canonical ensemble according to the
conjecture of arXiv:2505.21633. Here we study the ${\rm T}^{-1}_2$ correction for the M2 brane instanton wrapping $S^3/\mathbb Z_k\subset S^7/\mathbb Z_k$ dual to the leading non-perturbative large $N$ contribution to the ABJM free energy on 3-sphere. The 1-loop instanton M2 brane partition function was shown in arXiv:2307.14112 to match the localization prediction. The 2-loop computation is complicated by the presence of bosonic and fermionic zero modes, which requires to integrate over the collective coordinates. The collective-coordinate Jacobian provides a non-trivial ${\rm T}^{-1}_2$ contribution in addition to the one of the quartic interaction vertex in the M2 brane action. We find that for $k>2$ the the quartic vertex contribution is cancelled by the Jacobian contribution for the choice of the separation of the fermionic zero modes from the
quantum fields under which these form separate supermultiplets. Assuming the supersymmetric prescription for the paired zero modes we show that the cancellation persists after integration over the collective coordinates. Thus the ${\rm T}^{-1}_2$ correction vanishes, in agreement with the grand-canonical ensemble prediction. For $k=1$ the quartic vertex contribution reduces to equation-of-motion terms as in the case of ${\rm AdS}_3 \subset { \rm AdS}_7 \times S^4$ in arXiv:2511.22306 which is related to $S^3 \subset {\rm AdS}_4 \times S^7$ case by a formal analytic continuation.
Read the full abstract ▾