Abstract
For streaming speech recognition, a Transformer-based encoder has been widely used with block processing. Although many studies addressed improving emission latency of transducers, little work has been explored for improving encoding latency of the block processing. We seek to reduce latency by frequently emitting a chunk with a small shift rather than scarce large-chunk emissions, resulting in higher computational costs. To efficiently compute with the small chunk shift, we propose a new encoder, Spiralformer, tailored for block processing by combining layer dropping and early exiting. We skip layer computation in a cyclic manner and shift the computed layer in each block spirally, which completes computation for all the layers over the block processing. Experimentally, we observed that our method achieved 21.6% reduction in the averaged token emission delay in Librispeech, and 7.0% in CSJ, compared with the baseline with similar computational cost and word error rates.

Abstract
The widespread availability of cellular devices introduces new threat vectors that allow users or attackers to bypass security policies and physical barriers and bring unauthorized devices into sensitive areas. These threats can arise from user non-compliance or deliberate actions aimed at data exfiltration/infiltration via hidden devices, drones, etc. We identify a critical gap in this context: the absence of low-latency systems for high-quality and instantaneous monitoring of cellular transmissions. Such low-latency systems are crucial to allow for timely detection, decision (e.g., geofencing or localization), and disruption of unauthorized communication in sensitive areas. Operator-based monitoring systems, built for purposes such as people counting or tracking, lack real-time capability, require cooperation across multiple operators, and thus are hard to deploy. Operator-independent monitoring approaches proposed in the literature either lack low-latency capabilities or do not scale. We propose LTag, the first low-latency, operator-independent and scalable system designed to monitor cellular connections across all operators prior to any user data transmission. LTag consists of several downlink sniffers and a distributed network of uplink sniffers that measure both downlink protocol information and uplink signal characteristics at multiple locations to gain a detailed spatial image of uplink signals. LTag aggregates the recorded information, processes it, and provides a decision about the connection all prior to connection establishment of a UE. To evaluate LTag, we deployed it in the context of geofencing, where LTag was able to determine if the signals originate from inside or outside of an area within 2.3 ms of the initial base station-to-device message, therefore enabling prompt and targeted suppression of communication before any user data was transmitted.

Abstract
The learning crisis in the Latin American region (i.e., higher rates of students not reaching basic competencies at secondary level) is worrying, particularly post-pandemic given the stronger role of inequality behind achievement. Within this scenario, the concept of student academic resilience (SAR), students who despite coming from disadvantaged backgrounds reach good performance levels, and an analysis of its determinants, are policy relevant. In this paper, using advancements on explainable machine learning methods (the SHAP method) and relying on PISA 2022 data for 9 countries from the region, I identify leading factors behind SAR using diverse indicators. I find that household inputs (books and digital devices), gender, homework, repetition and work intensity are leading factors for one indicator of academic resilience, whereas for other indicator leading drives fall into the school domain: school size, the ratio of PC connected to the internet, STR and teaching quality proxied by certified teachers and professional development rates and school type (private school). Also, I find negative associations of SAR with the length of school closures and barriers for remote learning during the pandemic. The paper's findings adds to the scare regional literature as well as they contribute to future policy designs where key features behind SAR can be used to lift disadvantaged students from lower achievement groups towards being academic resilient.

Abstract
We present a data-driven framework for assessing the attack resilience of linear time-invariant systems against malicious false data injection sensor attacks. Based on the concept of sparse observability, data-driven resilience metrics are proposed. First, we derive a data-driven necessary and sufficient condition for assessing the system's resilience against sensor attacks, using data collected without any attacks. If we obtain attack-free data that satisfy a specific rank condition, we can exactly evaluate the attack resilience level even in a model-free setting. We then extend this analysis to a scenario where only poisoned data are available. Given the poisoned data, we can only conservatively assess the system's resilience. In both scenarios, we also provide polynomial-time algorithms to assess the system resilience under specific conditions. Finally, numerical examples illustrate the efficacy and limitations of the proposed framework.

Abstract
Large-scale international collaborations such as ATLAS rely on globally distributed workflows and data management to process, move, and store vast volumes of data. ATLAS's Production and Distributed Analysis (PanDA) workflow system and the Rucio data management system are each highly optimized for their respective design goals. However, operating them together at global scale exposes systemic inefficiencies, including underutilized resources, redundant or unnecessary transfers, and altered error distributions. Moreover, PanDA and Rucio currently lack shared performance awareness and coordinated, adaptive strategies. This work charts a path toward co-optimizing the two systems by diagnosing data-management pitfalls and prioritizing end-to-end improvements. With the observation of spatially and temporally imbalanced transfer activities, we develop a metadata-matching algorithm that links PanDA jobs and Rucio datasets at the file level, yielding a complete, fine-grained view of data access and movement. Using this linkage, we identify anomalous transfer patterns that violate PanDA's data-centric job-allocation principle. We then outline mitigation strategies for these patterns and highlight opportunities for tighter PanDA-Rucio coordination to improve resource utilization, reduce unnecessary data movement, and enhance overall system resilience.

Abstract
Although large language models (LLMs) have revolutionized natural language processing capabilities, their practical implementation as autonomous multi-agent systems (MAS) for industrial problem-solving encounters persistent barriers. Conventional MAS architectures are fundamentally restricted by inflexible, hand-crafted graph topologies that lack contextual responsiveness, resulting in diminished efficacy across varied academic and commercial workloads. To surmount these constraints, we introduce AMAS, a paradigm-shifting framework that redefines LLM-based MAS through a novel dynamic graph designer. This component autonomously identifies task-specific optimal graph configurations via lightweight LLM adaptation, eliminating the reliance on monolithic, universally applied structural templates. Instead, AMAS exploits the intrinsic properties of individual inputs to intelligently direct query trajectories through task-optimized agent pathways. Rigorous validation across question answering, mathematical deduction, and code generation benchmarks confirms that AMAS systematically exceeds state-of-the-art single-agent and multi-agent approaches across diverse LLM architectures. Our investigation establishes that context-sensitive structural adaptability constitutes a foundational requirement for high-performance LLM MAS deployments.