Search arXiv⌕ Search

arXiv subjects

Yu Mao

Publications and source records attributed to Yu Mao.

At least 19 recordsLinked to original sources
arXiv2026

How Far Are We From On-Device VLM Inference? A Measurement Study on Edge NPUs

Diverse optimizations spanning model architecture, inference runtime, and hardware acceleration have been proposed for on-device vision-language model (VLM) deployment on edge NPUs. Still, cross-layer coordination of VLM execution on edge NPUs remains largely underexplored. This paper presents an empirical measurement study across six representative VLMs, three real-world edge platforms, and ten multimodal benchmarks. Our evaluation drills down from end-to-end application-level latency and thermal traces down to low-level kernel-scale execution behaviors, challenging several widely held community assumptions. First, softmax dominates attention costs; even lookup-table-based softmax optimizations cannot eliminate pipeline bubbles that severely underutilize NPU matrix units. Second, theoretical speedups from linear attention vanish under real-NPU execution due to prohibitive tensor-rearrangement overhead. Third, no hardware backend universally dominates autoregressive decoding: attention-tensor-shaped parallelism dictates NPU efficiency and shifts the CPU-NPU performance balance. We further find task-specific Sweet Resolutions for VLMs, which deliver meaningful accuracy gains at only modest latency overhead. Complementing these findings, we provide actionable cross-layer recommendations for hardware architects, runtime developers, and model designers.

cs.PF↗
arXiv2026

Probe with Participation Trophies: Random-Reward RL as a Probe of LLM Capability

We connect the spurious-reward paradox to a model's reachability and propose random-reward reinforcement learning (RL) as a useful tool for the probing enterprise, addressing a decade-long debate over what probing performance actually reveals about a model. There are two prevailing explanations for the surprising finding that even random rewards can improve the performance of large language models (LLMs): one attributes the gains to particular mechanisms within RL training; the other to data contamination. Our results motivate a different view: spurious-reward RL can probe a model's reachability, or what further training can attain from its current state under specified constraints, beyond what is reflected in its current performance. Two OLMo checkpoints with the same accuracy on synthetic arithmetic (3.5%), for example, reach 8.5% and 55% in their best runs under the same correctness-rewarded RL. Examining OLMo checkpoints across pre-training and mid-training reveals three distinct regimes of training response: early on, RL produces little improvement even when correct answers are rewarded; later in pre-training, rewarding correct answers becomes effective while random rewards remain weak; and, upon entering mid-training, even random rewards can produce large gains. A similar ordering appears in a number-masked supervised fine-tuning (SFT) analysis of these checkpoints, suggesting that the pattern is not specific to a particular RL mechanism. Moreover, RL with random rewards offers a distinctive perspective on what training can make an LLM do, since its reward signal supplies no information about which answers are correct. By asking what training can attain without correctness feedback, it addresses the label-leakage side of a central problem in decodability-based probing: whether a successful probe reveals the model's capabilities or learns the task itself.

cs.CL↗
arXiv2026

AgentPProf: Semantic Profiler for Long Horizon AI Agents

AI agents increasingly orchestrate long-running activities with users, tools, and system resources for days and weeks. To improve agent quality, safety, and cost efficiency, developers need to determine where failures happen, what triggers unsafe effects, and which tasks consume the most budget, then optimize those tasks. In systems software, profiling answers similar questions by aggregating resource consumption and attributing it to responsible code paths to identify hotspots. Yet existing agent observability tools focus on per-execution debugging and tracing rather than cross-run, long term profiling, making these questions difficult to answer at scale. Agent observability needs profiling, not only debugging, but profiling agents is challenging: the responsible entities are task intent like diagnose authentication, compare branches rather than code paths, and lack stable identifiers for aggregation. We propose a semantic operation stack model that adapts profiling to agent trajectories. Uniform operations represent all activities, and operation stacks replace the runtime call stack, enabling hierarchical attribution at different granularities. We observe that an agent's task occupies a contiguous span and decomposes into subtasks, so we introduce recursive operation segmentation, which recursively splits trajectories at task boundaries. AgentPProf is a profiler that aggregates agent trajectories into pprof-compatible profiles, enabling flame graph visualization and analysis. AgentPProf reaches 0.764 $B^3$ F1 against human annotations on CodeTraceBench. On three problem-localization benchmarks, the profile raises MAP by up to 56%, demonstrating that it effectively attributes resources, locates problems, and helps optimize token cost at practical profiling cost. AgentPProf is available at https://github.com/eunomia-bpf/agentsight.

cs.AI↗
arXiv2026

LLM Agent Capabilities Should Follow Task Intent and Context Source

LLM agents take real actions, including executing code, modifying files, calling services, and delegating tasks, driven by context sources: user requests, tool results, documents, shell outputs, Skill and MCP instructions, memory. Unlike traditional systems, where capability is predefined, the least-privilege capability an agent needs is dynamic, depending on its task intent: what it wants to do and how. This creates a security and safety challenge: all inputs enter one shared planning channel with equal influence, by adversarial injection or accidental scope widening, yet a user's explicit request and a document's extracted text carry different trust and should not share authority to choose a destination or widen access. Existing defenses constrain operations and information flow; we study task-scoped, multi-source authority composition. We argue that agent capabilities should be scoped to the current task intent, not a sandbox or session lifetime, and that no single context source defines a complete capability. IntentCap composes capabilities from four such sources: user intent, workflow instructions, tool schemas, and runtime environment, with field-level ownership and monotonic narrowing. Each source contributes specific fields, none can fill another's, and the lease only narrows the user's authorized authority, never widens it. IntentCap uses an LLM to generate short-lived leases from these sources, validated by a deterministic checker before any side effect commits and enforced by tool- and OS-level information flow policies. Evaluation shows IntentCap blocks tested violations without rejecting benign actions, each source boundary is independently necessary, and the checker generalizes across tool, execution, placement, and delegation boundaries.

cs.CR↗
arXiv2026

FFSlim: An Efficient and Lightweight Format for Multi-modal Data Storage and Retrieval

With the rapid expansion of large-scale media-text corpora, multi-modal datasets increasingly require efficient storage and retrieval. Existing formats such as Files, TDP, and FFRecord work adequately for uni-modal data but expose fundamental limitations in multi-modal settings, including storage redundancy, massive small-file overheads, cache-unfriendly layouts, and heavy index structures. These issues jointly inflate storage and memory usage and make I/O the dominant bottleneck in real training workloads. We present FFSlim, a lightweight format for storing and retrieving multi-modal data. FFSlim improves storage efficiency and loading throughput through three components: a unified file format that removes media duplication and avoids small-file proliferation; an adaptive retrieval mechanism that enables low-overhead pair-level access and accelerates repeated media loading; and a redundancy detection and aggregation module that converts existing datasets into the FFSlim layout. The experimental results demonstrate that FFSlim achieves 2.07x and 8.26x higher data loading and write throughput on average than the strongest baseline, with minimal storage and index overhead. Consequently, these underlying I/O accelerations enable FFSlim to reduce end-to-end training time by 5.36%-14.18% across seven diverse multi-modal models.

cs.PF↗
arXiv2026

A Practical and Privacy-Preserving Framework for Real-World Large Language Model Services

Large language models (LLMs) have demonstrated exceptional capabilities in text understanding and generation, and they are increasingly being utilized across various domains to enhance productivity. However, due to the high costs of training and maintaining these models, coupled with the fact that some LLMs are proprietary, individuals often rely on online AI as a Service (AIaaS) provided by LLM companies. This business model poses significant privacy risks, as service providers may exploit users' trace patterns and behavioral data. In this paper, we propose a practical and privacy-preserving framework that ensures user anonymity by preventing service providers from linking requests to the individuals who submit them. Our framework is built on partially blind signatures, which guarantee the unlinkability of user requests. Furthermore, we introduce two strategies tailored to both subscription-based and API-based service models, ensuring the protection of both users' privacy and service providers' interests. The framework is designed to integrate seamlessly with existing LLM systems, as it does not require modifications to the underlying architectures. Experimental results demonstrate that our framework incurs minimal computation and communication overhead, making it a feasible solution for real-world applications.

cs.CR↗
arXiv2026

Do Models Hear Like Us? Probing the Representational Alignment of Audio LLMs and Naturalistic EEG

Audio Large Language Models (Audio LLMs) have demonstrated strong capabilities in integrating speech perception with language understanding. However, whether their internal representations align with human neural dynamics during naturalistic listening remains largely unexplored. In this work, we systematically examine layer-wise representational alignment between 12 open-source Audio LLMs and Electroencephalogram (EEG) signals across 2 datasets. Specifically, we employ 8 similarity metrics, such as Spearman-based Representational Similarity Analysis (RSA), to characterize within-sentence representational geometry. Our analysis reveals 3 key findings: (1) we observe a rank-dependence split, in which model rankings vary substantially across different similarity metrics; (2) we identify spatio-temporal alignment patterns characterized by depth-dependent alignment peaks and a pronounced increase in RSA within the 250-500 ms time window, consistent with N400-related neural dynamics; (3) we find an affective dissociation whereby negative prosody, identified using a proposed Tri-modal Neighborhood Consistency (TNC) criterion, reduces geometric similarity while enhancing covariance-based dependence. These findings provide new neurobiological insights into the representational mechanisms of Audio LLMs.

cs.SD↗
arXiv2025

Open Homomorphisms between $m$-step Solvable Galois Groups Compatible with the Cyclotomic Characters

In \cite{Ho3}, Hoshi proved that open homomorphisms between solvably closed Galois groups of number fields which are compatible with the cyclotomic characters arise from field embeddings. In this paper, we will prove an $m$-step solvable version of Hoshi's result. More precisely, if $K$ and $L$ are number fields, we will prove that given an open homomorphism between the maximal $m+3$-step solvable Galois groups of $K$ and $L$, where $m \geq 2$, and the induced open homomorphism between the corresponding maximal $m$-step solvable Galois groups, then the latter arises from a field embedding if and only if the open homomorphism between the $m+3$-step solvable (and hence also the $m$-step solvable) Galois groups is compatible with the cyclotomic characters of $K$ and $L$.

math.NT↗
arXiv2025

Skeletons Matter: Dynamic Data Augmentation for Text-to-Query

The task of translating natural language questions into query languages has long been a central focus in semantic parsing. Recent advancements in Large Language Models (LLMs) have significantly accelerated progress in this field. However, existing studies typically focus on a single query language, resulting in methods with limited generalizability across different languages. In this paper, we formally define the Text-to-Query task paradigm, unifying semantic parsing tasks across various query languages. We identify query skeletons as a shared optimization target of Text-to-Query tasks, and propose a general dynamic data augmentation framework that explicitly diagnoses model-specific weaknesses in handling these skeletons to synthesize targeted training data. Experiments on four Text-to-Query benchmarks demonstrate that our method achieves state-of-the-art performance using only a small amount of synthesized data, highlighting the efficiency and generality of our approach and laying a solid foundation for unified research on Text-to-Query tasks. We release our code at https://github.com/jjjycaptain/Skeletron.

cs.CL↗
arXiv2025

The 6-step Solvable Mono-anabelian Reconstruction of Abelian Number Fields

In this paper, we develop a new method to reconstruct an abelian number field $K$ from the maximal $6$-step solvable quotient of $G_K$ group- theoretically. The new aspect of this paper is that the results in this paper are independent from the bi-anabelian results proved proven by Saidi and Tamagawa in [ST22].

math.NT↗
arXiv2025

The m-step Solvable Mono-anabelian Geometry of Number Fields

The goal of this paper is to develop a group-theoretic algorithm, to reconstruct a number field (together with its maximal m-step solvable ex- tension for some positive integer m \geq 3) from the maximal m+9-step solv- able quotient of its absolute Galois group. If K is an imaginary quadratic field or Q, we establish a group-theoretic reconstruction algorithm of K from the maximal 6-step solvable quotient of its absolute Galois group.

math.NT↗
arXiv2025

Catalogue of chiral phonon materials

Chiral phonons, circularly polarized lattice vibrations carrying intrinsic angular momentum, offer unprecedented opportunities for controlling heat flow, manipulating quantum states through spin-phonon coupling, and realizing exotic transport phenomena. Despite their fundamental importance, a universal framework for identifying and classifying these elusive excitations has remained out of reach. Here, we address this challenge by establishing a comprehensive symmetry-based theory that systematically classifies the helicity and the velocity-angular momentum tensor underlying phonon magnetization in thermal transport across all 230 crystallographic space groups. Our approach, grounded in fundamental representations of phononic angular momentum, reveals three distinct classes of crystals: achiral crystals with vanishing angular momentum, chiral crystals with s-wave helicity, and achiral crystals exhibiting higher-order helicity patterns beyond the s-wave. By performing high-throughput computations and symmetry analysis of the dynamical matrices for 11614 crystalline compounds, we identified 2738 materials exhibiting chiral phonon modes and shortlisted the 170 most promising candidates for future experimental investigation. These results are compiled into an open-access Chiral Phonon Materials Database website, enabling rapid screening for materials with desired chiral phonon properties. Our theoretical framework transcends phonons--it provides a universal paradigm for classifying chiral excitations in crystalline lattices, from magnons to electronic quasiparticles.

cond-mat.mtrl-sci↗
arXiv2025

Easz: An Agile Transformer-based Image Compression Framework for Resource-constrained IoTs

Neural image compression, necessary in various machine-to-machine communication scenarios, suffers from its heavy encode-decode structures and inflexibility in switching between different compression levels. Consequently, it raises significant challenges in applying the neural image compression to edge devices that are developed for powerful servers with high computational and storage capacities. We take a step to solve the challenges by proposing a new transformer-based edge-compute-free image coding framework called Easz. Easz shifts the computational overhead to the server, and hence avoids the heavy encoding and model switching overhead on the edge. Easz utilizes a patch-erase algorithm to selectively remove image contents using a conditional uniform-based sampler. The erased pixels are reconstructed on the receiver side through a transformer-based framework. To further reduce the computational overhead on the receiver, we then introduce a lightweight transformer-based reconstruction structure to reduce the reconstruction load on the receiver side. Extensive evaluations conducted on a real-world testbed demonstrate multiple advantages of Easz over existing compression approaches, in terms of adaptability to different compression levels, computational efficiency, and image reconstruction quality.

eess.IV↗
arXiv2025

Lossless Compression of Large Language Model-Generated Text via Next-Token Prediction

As large language models (LLMs) continue to be deployed and utilized across domains, the volume of LLM-generated data is growing rapidly. This trend highlights the increasing importance of effective and lossless compression for such data in modern text management systems. However, compressing LLM-generated data presents unique challenges compared to traditional human- or machine-generated content. Traditional machine-generated data is typically derived from computational processes or device outputs, often highly structured and limited to low-level elements like labels or numerical values. This structure enables conventional lossless compressors to perform efficiently. In contrast, LLM-generated data is more complex and diverse, requiring new approaches for effective compression. In this work, we conduct the first systematic investigation of lossless compression techniques tailored specifically to LLM-generated data. Notably, because LLMs are trained via next-token prediction, we find that LLM-generated data is highly predictable for the models themselves. This predictability enables LLMs to serve as efficient compressors of their own outputs. Through extensive experiments with 14 representative LLMs and 8 LLM-generated datasets from diverse domains, we show that LLM-based prediction methods achieve remarkable compression rates, exceeding 20x, far surpassing the 3x rate achieved by Gzip, a widely used general-purpose compressor. Furthermore, this advantage holds across different LLM sizes and dataset types, demonstrating the robustness and practicality of LLM-based methods in lossless text compression under generative AI workloads.

cs.LG↗
arXiv2025

WISE: A Framework for Gigapixel Whole-Slide-Image Lossless Compression

Whole-Slide Images (WSIs) have revolutionized medical analysis by presenting high-resolution images of the whole tissue slide. Despite avoiding the physical storage of the slides, WSIs require considerable data volume, which makes the storage and maintenance of WSI records costly and unsustainable. To this end, this work presents the first investigation of lossless compression of WSI images. Interestingly, we find that most existing compression methods fail to compress the WSI images effectively. Furthermore, our analysis reveals that the failure of existing compressors is mainly due to information irregularity in WSI images. To resolve this issue, we developed a simple yet effective lossless compressor called WISE, specifically designed for WSI images. WISE employs a hierarchical encoding strategy to extract effective bits, reducing the entropy of the image and then adopting a dictionary-based method to handle the irregular frequency patterns. Through extensive experiments, we show that WISE can effectively compress the gigapixel WSI images to 36 times on average and up to 136 times.

eess.IV↗
arXiv2025

When Compression Meets Model Compression: Memory-Efficient Double Compression for Large Language Models

Large language models (LLMs) exhibit excellent performance in various tasks. However, the memory requirements of LLMs present a great challenge when deploying on memory-limited devices, even for quantized LLMs. This paper introduces a framework to compress LLM after quantization further, achieving about 2.2x compression ratio. A compression-aware quantization is first proposed to enhance model weight compressibility by re-scaling the model parameters before quantization, followed by a pruning method to improve further. Upon this, we notice that decompression can be a bottleneck during practical scenarios. We then give a detailed analysis of the trade-off between memory usage and latency brought by the proposed method. A speed-adaptive method is proposed to overcome it. The experimental results show inference with the compressed model can achieve a 40% reduction in memory size with negligible loss in accuracy and inference speed.

cs.CL↗
arXiv2024

BAHOP: Similarity-based Basin Hopping for A fast hyper-parameter search in WSI classification

Pre-processing whole slide images (WSIs) can impact classification performance. Our study shows that using fixed hyper-parameters for pre-processing out-of-domain WSIs can significantly degrade performance. Therefore, it is critical to search domain-specific hyper-parameters during inference. However, searching for an optimal parameter set is time-consuming. To overcome this, we propose BAHOP, a novel Similarity-based Basin Hopping optimization for fast parameter tuning to enhance inference performance on out-of-domain data. The proposed BAHOP achieves 5\% to 30\% improvement in accuracy with $\times5$ times faster on average.

cs.CV↗