Showing 1–50 of 140 results for author: Tsai, M

Spatial Message Passing in Language Space for Pathology Image Interpretation

Authors: Jing-Cheng Yang, Hao-Jung Wang, Jinhao Du, Yang Hu, Ming-shan Tsai, Jens Rittscher, Bin Li

Abstract: Multimodal Large Language Models (MLLMs) can generate pathological descriptions from histological images, but gigapixel Whole Slide Images (WSIs) exceed their visual context limits. The standard tiling workaround makes WSIs tractable yet severs the tissue neighborhoods that define tumor-stroma interfaces and morphology. We introduce Spatial Language Message Passing (SLMP), a framework that perform… ▽ More Multimodal Large Language Models (MLLMs) can generate pathological descriptions from histological images, but gigapixel Whole Slide Images (WSIs) exceed their visual context limits. The standard tiling workaround makes WSIs tractable yet severs the tissue neighborhoods that define tumor-stroma interfaces and morphology. We introduce Spatial Language Message Passing (SLMP), a framework that performs spatial reasoning entirely in language space, human-readable by construction. SLMP represents a WSI region as a spatial text graph: tiles are nodes initialized with MLLM descriptions, and edges encode spatial adjacency. For each tile, an LLM refines its description by integrating language messages from adjacent tiles under a shared aggregation policy that, on the tile grid, acts as an adaptive local kernel operating on text rather than learned embeddings. This policy is an inspectable prompt that can be refined from model-observed tissue phenotypes via textual gradients, enabling automatic semantic optimization from local cellular context to broader tissue morphology without fine-tuning MLLM weights. On representative HER2 and CAMELYON16 regions, SLMP improves tile-level tumor description accuracy in settings spanning general-purpose and pathology-specialized backbones, with gains of +3.3 to +19.6 percentage points. Random-neighbor ablations confirm that these gains stem from spatial context rather than additional text alone, and inspecting the optimized policies reveals interpretable, tissue-specific decision rules. Besides, without any weight updates or fine-tuning the backbone MLLM, SLMP substantially improves general-purpose MLLMs and narrows its gap to pathology-specialized counterparts, offering a transparent and flexible mechanism for incorporating spatial reasoning into MLLM-based pathology analysis. △ Less

Submitted 14 August, 2026; originally announced August 2026.

Comments: Accepted at MICCAI 2026 Workshop (Oral)

Triton for MTIA: Bridging the Programming Model Gaps for Custom AI Accelerators

Authors: Haishan Zhu, Domi Yan, Michael Levesque-Dion, Changxu Zhang, Mitch Gamburg, Kirsten Lee, Giancarlo Colmenares, Aditya Bhagwat, Arnab De, Markus Le Roux, Victor Perez Carrasco, Xin Tong, Will Cromar, Simran Barnwal, Andrew Uderian, Sridhar Gopinath, Jan Szczepaniec, Daniel Neilson, Blaine Burton Rister, Jordan Fix, Jazlyn Li, Zejun Huang, Lite Ye, Nan Zhang, Xinchen Guo , et al. (18 additional authors not shown)

Abstract: The rapid growth in machine learning workloads has fueled the proliferation of custom accelerator architectures. Designed from the ground up, these accelerators often expose programming models that are distinct from GPUs. While hyperscalers and AI chip startups continue to innovate in this space, achieving broad operator coverage to support diverse models remains a major challenge. Additionally, a… ▽ More The rapid growth in machine learning workloads has fueled the proliferation of custom accelerator architectures. Designed from the ground up, these accelerators often expose programming models that are distinct from GPUs. While hyperscalers and AI chip startups continue to innovate in this space, achieving broad operator coverage to support diverse models remains a major challenge. Additionally, an easy-to-use, high-level kernel programming language is important for rapid iteration of models and kernels. Triton, together with TorchInductor, addresses these issues on GPUs, but its viability on accelerators with different programming models has yet to be established. In this work, we present the first production-scale application of Triton on a custom ML accelerator, MTIA-2i, developed by Meta. To support MTIA-2i, we develop a new compiler backend that targets it, introduce enhancements to TorchInductor code generation, and propose minimal language extensions that expose MTIA-specific architectural features. We demonstrate that Triton-MTIA kernels achieve performance competitive with expert-tuned C++ implementations. Leveraging these development efficiency gains, we successfully deployed manually written and Inductor-generated Triton kernels in production across approximately 60 different model types, accounting for 50% of layers and 47% of non-GEMM execution time for these models. Our results provide compelling evidence that DSLs like Triton can bridge the programming model gaps between ML frameworks, kernels, and custom accelerators, enabling rapid innovation and efficient deployment at scale. △ Less

Submitted 12 August, 2026; v1 submitted 31 July, 2026; originally announced August 2026.

Comments: 12 pages, 12 figures, to be published in IEEE Micro

Absent, Not Faint: Fisher-Information Limits and a Logarithmic Measurement-Design Cure for Passive Characterization of Coherent Qubit Noise

Authors: Yi Pan, Meng Hsiu Tsai, Weihang You, Hanqi Jiang, Junhao Chen, Wei Zhang, Isaac Lyngaas, Yingfeng Wang, Tianming Liu

Abstract: Calibrating a quantum processor means estimating error parameters, and estimation theory usually assumes a parameter hard to estimate is faint: its signal is weak but present, so more repetitions or a richer model will recover it. This assumption fails for a leading hardware fault. A coherent over-rotation is a small systematic gate miscalibration. Measured through the cheapest data a device retur… ▽ More Calibrating a quantum processor means estimating error parameters, and estimation theory usually assumes a parameter hard to estimate is faint: its signal is weak but present, so more repetitions or a richer model will recover it. This assumption fails for a leading hardware fault. A coherent over-rotation is a small systematic gate miscalibration. Measured through the cheapest data a device returns--one fixed-basis histogram--it is not faint but absent: to first order it leaves the distribution unchanged, indistinguishable from a compensating stochastic error, exactly as two numbers cannot be separated from their sum. For commuting single- and two-qubit transverse over-rotations, with known support on the canonical input, the histogram's Fisher information is singular along the fault's direction at zero angle, its Cramer-Rao bound is infinite, and no finite-variance, locally unbiased estimator recovers it. At a generic nonzero angle the degeneracy partly lifts; beyond four qubits it clears entirely, leaving conditioning, not absence, as the obstruction. The cure is a richer measurement, not a richer model: a fixed, logarithmically small set of extra settings makes every such fault visible. Visibility alone is not enough. The sampling cost is set by conditioning, not coverage, through a floor whose complete-family closed form is exponentially small in the qubit count. We prove the impossibility and cure, confirm both in exact simulation, show conditioning predicts recovery error across hundreds of designs, and observe a 3-5x bias gap on IBM Heron hardware as a consistency check. Non-commuting faults and unknown support remain open. △ Less

Submitted 22 July, 2026; originally announced July 2026.

LLMs and Agentic AI Systems for Smart Grids: A Tutorial on Architectures and Applications

Authors: Daniela Rojas, Abdulwahab Albassam, Aidan G. Leung, Jett Ngo, Ryan Luo, Peter R. Quawas, Junpyung Kim, Kangkai Liang, Mansi Nanavati, Jonathan Mai, Meng-Chi Tsai, Yun-Tong Tsai, Yize Chen, Yuanyuan Shi

Abstract: Large language models (LLMs) and agentic AI systems have evolved from natural language tasks to using external tools to plan, retrieve, and act in technical domains. In smart grids, recent work applies agentic schemes to forecasting, optimization, and control, wrapping trusted solvers behind language interfaces and orchestrating multi-step workflows. The literature lacks a unified approach to desi… ▽ More Large language models (LLMs) and agentic AI systems have evolved from natural language tasks to using external tools to plan, retrieve, and act in technical domains. In smart grids, recent work applies agentic schemes to forecasting, optimization, and control, wrapping trusted solvers behind language interfaces and orchestrating multi-step workflows. The literature lacks a unified approach to designing and evaluating such systems. LLMs can produce numerically plausible yet physically infeasible outputs, evaluation protocols vary across tasks, and the boundary between what the model should and should not compute is implicit. This paper presents a solver-grounded design principle: a numerical result is reported only when it originates from a trusted tool and passes explicit verification. We review the building blocks of LLM and agentic AI systems for power systems: prompting strategies and agentic architectures. We instantiate the principle in four case studies: wind power forecasting, EV charging scheduling, power flow analysis, and contingency diagnosis, each comparing an LLM-only baseline against its solver-grounded counterpart on identical data and metrics. EVAgent reproduces the CVXPY optimum while reducing LLM-only unmet energy by 7.5-9.5x, and GridDebugAgent repairs 17/39 contingency cases while reducing total violations by 52.3%. We propose a four-group evaluation framework spanning task utility, solver-grounded correctness, faithfulness and safe failure, and cost and latency. A consistent division of labor emerges: the agentic system reliably orchestrates, retrieves, and explains, while trusted tools compute and a verification gate decides what is reported. △ Less

Submitted 20 July, 2026; originally announced July 2026.

Comments: 28 pages, 11 figures, 6 tables; plus supplementary material. Review/tutorial article

Local Certification of Vertex and Edge Connectivity

Authors: Yi-Jun Chang, Yi-Xuan Lee, Meng-Tsung Tsai

Abstract: Local certification is a framework for verifying global graph properties using only local information. In this model, a prover assigns short labels, called certificates, to the vertices of a graph. Each vertex then exchanges certificates with its neighbors and performs a purely local check to determine whether the graph satisfies the desired property. This line of research has led to efficient cer… ▽ More Local certification is a framework for verifying global graph properties using only local information. In this model, a prover assigns short labels, called certificates, to the vertices of a graph. Each vertex then exchanges certificates with its neighbors and performs a purely local check to determine whether the graph satisfies the desired property. This line of research has led to efficient certification schemes for a broad range of graph classes, including minor-closed families, topological graph classes, and graphs defined by forbidden subgraphs. In this paper, we study the local certification of graph connectivity. Prior work by Bousquet, Feuilloley, and Pierron (JPDC 2024) showed that $2$-vertex-connectivity, $2$-edge-connectivity, and $3$-vertex-connectivity admit $O(\log n)$-bit certificates, leveraging structural characterizations such as ear decompositions. We go substantially beyond these cases and investigate general $k$-vertex-connectivity and $k$-edge-connectivity. We develop new approaches that exploit connections between connectivity and combinatorial structures, including branchings, Eulerian subgraphs, and independent spanning trees. For $k$-edge-connectivity, we obtain an $O_k(\log n)$-bit certification scheme and prove a matching $Ω_k(\log n)$ lower bound for every $k\ge 3$. The lower bound also applies to $k$-vertex-connectivity. For $k$-vertex-connectivity, we obtain $\tilde{O}_k(\sqrt{n})$-bit certificates for every $k$ under a conjecture of Itai and Zehavi. We further show that, for $k=2$, the logarithmic barrier can be broken on sparse graph classes: $2$-edge-connectivity admits constant-size certificates in bounded-expansion graphs, and $2$-vertex-connectivity admits constant-size certificates in bounded-degree graphs. In contrast, for $2$-vertex-connectivity in general graphs, we prove an $Ω(\log(\log^\ast n))$-bit lower bound. △ Less

Submitted 15 July, 2026; originally announced July 2026.

Comments: Abstract shortened to meet arXiv requirements

TokenMinds: Pretrained User Tokens and Embeddings for User Understanding in Large Recommender Systems

Authors: Qingyun Liu, Bo Yan, Yang Liu, Yuji Roh, Ekansh Sharma, Likang Yin, Emma Olowo, Min-hsuan Tsai, Yuxuan Li, Diego Uribe, Saksham Aggarwal, Siqi Wu, Yuan Hao, Vikas Kedigehalli, Lukasz Heldt, Lichan Hong, Li Wei, Xinyang Yi

Abstract: User modeling in industrial recommender systems typically produces dense embeddings, which suffer from representational constraints inherent to fixed-dimensional vectors. An emerging alternative for discrete user representation -- using LLMs to generate text-based user tokens -- captures topical co-occurrences rather than deep sequential behavior dynamics and produces outputs that are difficult to… ▽ More User modeling in industrial recommender systems typically produces dense embeddings, which suffer from representational constraints inherent to fixed-dimensional vectors. An emerging alternative for discrete user representation -- using LLMs to generate text-based user tokens -- captures topical co-occurrences rather than deep sequential behavior dynamics and produces outputs that are difficult to ground to item attributes. Meanwhile, Semantic ID (SID) based item tokenization has proven effective for improving generalization in generative recommendation, yet discrete SID-based representations for users remain largely unexplored. We propose TokenMinds, an industrial-scale system that extends the PLUM framework from item retrieval to user modeling, generating both discrete SID-based user tokens and dense user embeddings via an encoder-decoder architecture adapted from pre-trained LLMs. This dual-output design provides the complementary benefits of discrete, semantically grounded user representations while maintaining compatibility with existing downstream models that rely on dense embeddings. Additionally, the shared SID vocabulary naturally extends to cross-scenario modeling: by unifying long-form and short-form video behaviors into a single model, we substantially reduce training and serving costs. We validate TokenMinds through extensive offline experiments and live launches on multiple YouTube surfaces, served on full user traffic (billions of users) via an asynchronous infrastructure that decouples representation generation from downstream scoring. Focusing on ranking as the primary downstream use case, our results confirm the practical viability of SID-based user tokens at industrial scale and demonstrate that tokens and dense embeddings provide complementary value across different production ranking systems. △ Less

Submitted 23 June, 2026; originally announced June 2026.

Quantization as a Malicious Task: Removing Quantization-Conditioned Backdoors via Task Arithmetic

Authors: Kaihsun Yang, Min-Yan Tsai, Chia-Mu Yu

Abstract: Model quantization is widely adopted to reduce memory usage and inference cost when deploying deep neural networks on resource-constrained devices. However, recent studies have revealed a new security threat known as Quantization-Conditioned Backdoors (QCBs), where a model behaves normally in full precision but activates malicious behavior only after quantization. Existing defenses typically modif… ▽ More Model quantization is widely adopted to reduce memory usage and inference cost when deploying deep neural networks on resource-constrained devices. However, recent studies have revealed a new security threat known as Quantization-Conditioned Backdoors (QCBs), where a model behaves normally in full precision but activates malicious behavior only after quantization. Existing defenses typically modify quantization procedures or correct activation statistics, often introducing additional computational overhead or relying on specific quantization settings. Here, we present QVec, a parameter-space perspective for defending against QCBs. We observe that the weight difference between a full-precision model and its quantized counterpart encodes a structured behavioral shift, which can be interpreted as a malicious task vector rather than random quantization noise. Based on this insight, QVec counteracts this malicious direction through controlled parameter correction prior to deployment. QVec requires no retraining, no trigger samples, and only a single quantization pass to estimate the parameter shift, together with a lightweight hyperparameter search. Extensive experiments across image classification benchmarks and multiple Large Language Model (LLM) attack scenarios demonstrate that QVec consistently suppresses backdoor activation while preserving clean performance. △ Less

Submitted 18 June, 2026; originally announced June 2026.

Towards Compact Sign Language Translation: Frame Rate and Model Size Trade-offs

Authors: Kuanwei Chen, Mengfeng Tsai

Abstract: Sign Language Translation (SLT) converts sign language videos into spoken-language text, bridging communication between Deaf and hearing communities. Current gloss-free approaches rely on large encoder-decoder models, limiting deployment. We propose a compact 77M-parameter pipeline that couples MMPose skeletal pose extraction with a single linear projection into T5-small. By varying the input fram… ▽ More Sign Language Translation (SLT) converts sign language videos into spoken-language text, bridging communication between Deaf and hearing communities. Current gloss-free approaches rely on large encoder-decoder models, limiting deployment. We propose a compact 77M-parameter pipeline that couples MMPose skeletal pose extraction with a single linear projection into T5-small. By varying the input frame rate, we expose a practical efficiency trade-off: at 12 fps the model halves its sequence length, achieving a 75% reduction in encoder quadratic self-attention computational complexity while incurring only a modest BLEU-4 drop (9.53 vs. 10.06 at 24 fps on How2Sign). Our system is roughly 3x smaller than prior T5-base systems, demonstrating that a lightweight architecture can remain competitive without hierarchical encoders or large-scale models. △ Less

Submitted 10 May, 2026; originally announced May 2026.

Comments: 2 pages, 1 figure, 2 tables

POLARIS: A Sparse Radial Neutrino Telescope Design for the Pacific Ocean

Authors: Karolin Hymon, Alexander Chen, Meng-Xue Tsai, Wan-Ting Hseu, Tzu-Hsuan Su, Anatoli Fedynitch

Abstract: The cubic-kilometer neutrino telescopes have opened neutrino astronomy as an observational discipline. The recent detection of KM3-230213A, the highest-energy neutrino ever observed at ~220 PeV, as a near-horizontal muon track underscores that the ultra-high-energy frontier is accessed through horizontal directions where the Earth's opacity above ~100 TeV confines the observable sky to a narrow ba… ▽ More The cubic-kilometer neutrino telescopes have opened neutrino astronomy as an observational discipline. The recent detection of KM3-230213A, the highest-energy neutrino ever observed at ~220 PeV, as a near-horizontal muon track underscores that the ultra-high-energy frontier is accessed through horizontal directions where the Earth's opacity above ~100 TeV confines the observable sky to a narrow band around and above the horizon. Yet extending general-purpose detector architectures into this regime requires disproportionate increases in instrumentation, cost, and logistical complexity. A compelling alternative is to deploy specialized detectors that target this natural geometry. POLARIS (Pacific Ocean Large Area Radial Instrumented Sparse array) is a sparse planar deep-water Cherenkov array optimized for neutrino-induced muon tracks from horizontal directions in the multi-TeV to PeV regime. By rotating the conventional vertical string layout into a radial planar configuration, the detector presents maximal cross-section to horizontal tracks while naturally suppressing the down-going atmospheric background. With only 1100 optical modules, the five-arm design reaches point source and diffuse flux sensitivities at PeV energies competitive with detectors deploying several times more instrumentation. As a dedicated $ν_μ$ track detector, POLARIS provides the muon-flavor channel that tau-optimized experiments such as TAMBO and Trinity do not cover, enabling full flavor composition measurements from astrophysical sources. Using the Prometheus simulation framework, this study demonstrates that targeted sparse geometries can open new discovery space at the high-energy frontier at a fraction of the cost of general-purpose arrays. △ Less

Submitted 22 April, 2026; v1 submitted 14 April, 2026; originally announced April 2026.

Comments: Submission to JCAP

Incorporating circular economy policies into product supply chains using bilevel optimization -- A case study on coffee packaging

Authors: Paola Munoz Briones, Meng-Lin Tsai, Styliani Avraamidou

Abstract: Transitioning to a Circular Economy requires policies to drive sustainable practices. This study proposes a bilevel optimization framework to evaluate the combined use of carbon taxes and subsidies in promoting circular supply chains under varying budget levels. A case study of the coffee packaging supply chain with an Extended Producer Responsibility scenario is used to demonstrate this approach.… ▽ More Transitioning to a Circular Economy requires policies to drive sustainable practices. This study proposes a bilevel optimization framework to evaluate the combined use of carbon taxes and subsidies in promoting circular supply chains under varying budget levels. A case study of the coffee packaging supply chain with an Extended Producer Responsibility scenario is used to demonstrate this approach. The framework captures the hierarchical interaction between a regional government (upper level), which aims to minimize environmental impacts, and coffee companies (lower level), which seek to minimize costs. Two bilevel optimization problems are formulated based on two environmental objectives: (1) minimization of greenhouse gas (GHG) emissions, and (2) maximization of circularity. The model integrates mixed-integer linear programming (MILP) with life cycle assessment (LCA), techno-economic assessment (TEA) and circularity assessment. Results demonstrate that subsidies effectively drive supply chain shifts toward low-emission and high-circularity configurations, while carbon taxes alone have a more limited impact. Sensitivity analyses highlight the influence of key parameters, such as glass washing distance and loss rates, on policy effectiveness. Overall, the study provides a bilevel optimization framework with quantitative insights to support policy design for sustainable circular supply chains. △ Less

Submitted 2 April, 2026; originally announced April 2026.

Parametric Region Search: A Mixed-Integer Bilevel Optimization Problem Primal Heuristic

Authors: Meng-Lin Tsai, Parth Brahmbhatt, Styliani Avraamidou

Abstract: Bilevel optimization is a mathematical modeling formulation for hierarchical systems and two-player interactions, with wide-ranging applications in environmental, energy, and control engineering. Despite its utility, the mixed-integer bilevel optimization (MIBO) problem is exceptionally challenging to solve. While numerous exact and metaheuristic methods exist, the development of specialized prima… ▽ More Bilevel optimization is a mathematical modeling formulation for hierarchical systems and two-player interactions, with wide-ranging applications in environmental, energy, and control engineering. Despite its utility, the mixed-integer bilevel optimization (MIBO) problem is exceptionally challenging to solve. While numerous exact and metaheuristic methods exist, the development of specialized primal heuristics for MIBO, aimed at quickly identifying high-quality feasible solutions, remains an underexplored area. This paper introduces the Parametric Region Search (PRS), a new primal heuristic for MIBO. The PRS method leverages insights from multi-parametric optimization by iteratively exploring regions defined by the lower-level problem's critical regions. We formally define the MIBO structure and the necessary parametric region formulations, and then detail the proposed heuristic's initialization and iterative search mechanism. Computational results demonstrate that the PRS heuristic consistently locates high-quality primal solutions compared to established derivative-free metaheuristics, including DOMINO-COBYLA and DOMINO-ISRES. Furthermore, we illustrate how the PRS can be effectively integrated with other heuristics like DOMINO-COBYLA to enhance the overall solution discovery process for MIBO. △ Less

Submitted 31 March, 2026; originally announced April 2026.

Reward-Based Online LLM Routing via NeuralUCB

Authors: Ming-Hua Tsai, Phat Tran

Abstract: This study investigates the use of NeuralUCB for cost-aware large language model (LLM) routing. Existing routing approaches can be broadly grouped into supervised routing methods and partial-feedback methods, each with different tradeoffs in efficiency and adaptivity. We implement a NeuralUCB-based routing policy and evaluate it on RouterBench under a simulated online setting. Experimental results… ▽ More This study investigates the use of NeuralUCB for cost-aware large language model (LLM) routing. Existing routing approaches can be broadly grouped into supervised routing methods and partial-feedback methods, each with different tradeoffs in efficiency and adaptivity. We implement a NeuralUCB-based routing policy and evaluate it on RouterBench under a simulated online setting. Experimental results show that the proposed method consistently outperforms random and min-cost baselines in utility reward. Compared with the max-quality reference, our method achieves substantially lower inference cost while maintaining competitive reward. These findings suggest that NeuralUCB is a promising approach for cost-aware LLM routing, while also highlighting remaining challenges in action discrimination and exploration. △ Less

Submitted 31 March, 2026; originally announced March 2026.

MS2MetGAN: Latent-space adversarial training for metabolite-spectrum matching in MS/MS database search

Authors: Meng Tsai, Alexzander Dwyer, Estelle Nuckels, Yingfeng Wang

Abstract: Database search is a widely used approach for identifying metabolites from tandem mass spectra (MS/MS). In this strategy, an experimental spectrum is matched against a user-specified database of candidate metabolites, and candidates are ranked such that true metabolite-spectrum matches receive the highest scores. Machine-learning methods have been widely incorporated into database-search-based ide… ▽ More Database search is a widely used approach for identifying metabolites from tandem mass spectra (MS/MS). In this strategy, an experimental spectrum is matched against a user-specified database of candidate metabolites, and candidates are ranked such that true metabolite-spectrum matches receive the highest scores. Machine-learning methods have been widely incorporated into database-search-based identification tools and have substantially improved performance. To further improve identification accuracy, we propose a new framework for generating negative training samples. The framework first uses autoencoders to learn latent representations of metabolite structures and MS/MS spectra, thereby recasting metabolite-spectrum matching as matching between latent vectors. It then uses a GAN to generate latent vectors of decoy metabolites and constructs decoy metabolite-spectrum matches as negative samples for training. Experimental results show that our tool, MS2MetGAN, achieves better overall performance than existing metabolite identification methods. △ Less

Submitted 6 March, 2026; originally announced March 2026.

HeartAgent: An Autonomous Agent System for Explainable Differential Diagnosis in Cardiology

Authors: Shuang Zhou, Kai Yu, Song Wang, Wenya Xie, Zaifu Zhan, Meng-Han Tsai, Yuen-Hei Chung, Shutong Hou, Huixue Zhou, Min Zeng, Bhavadharini Ramu, Lin Yee Chen, Feng Xie, Rui Zhang

Abstract: Heart diseases remain a leading cause of morbidity and mortality worldwide, necessitating accurate and trustworthy differential diagnosis. However, existing artificial intelligence-based diagnostic methods are often limited by insufficient cardiology knowledge, inadequate support for complex reasoning, and poor interpretability. Here we present HeartAgent, a cardiology-specific agent system design… ▽ More Heart diseases remain a leading cause of morbidity and mortality worldwide, necessitating accurate and trustworthy differential diagnosis. However, existing artificial intelligence-based diagnostic methods are often limited by insufficient cardiology knowledge, inadequate support for complex reasoning, and poor interpretability. Here we present HeartAgent, a cardiology-specific agent system designed to support a reliable and explainable differential diagnosis. HeartAgent integrates customized tools and curated data resources and orchestrates multiple specialized sub-agents to perform complex reasoning while generating transparent reasoning trajectories and verifiable supporting references. Evaluated on the MIMIC dataset and a private electronic health records cohort, HeartAgent achieved over 36% and 20% improvements over established comparative methods, in top-3 diagnostic accuracy, respectively. Additionally, clinicians assisted by HeartAgent demonstrated gains of 26.9% in diagnostic accuracy and 22.7% in explanatory quality compared with unaided experts. These results demonstrate that HeartAgent provides reliable, explainable, and clinically actionable decision support for cardiovascular care. △ Less

Submitted 11 March, 2026; originally announced March 2026.

Comments: 26 pages, 7 figures

SpecLoop: An Agentic RTL-to-Specification Framework with Formal Verification Feedback Loop

Authors: Fu-Chieh Chang, Yu-Hsin Yang, Hung-Ming Huang, Yun-Chia Hsu, Yin-Yu Lin, Ming-Fang Tsai, Chun-Chih Yang, Pei-Yuan Wu

Abstract: RTL implementations frequently lack up-to-date or consistent specifications, making comprehension, maintenance, and verification costly and error-prone. While prior work has explored generating specifications from RTL using large language models (LLMs), ensuring that the generated documents faithfully capture design intent remains a major challenge. We present SpecLoop, an agentic framework for RT… ▽ More RTL implementations frequently lack up-to-date or consistent specifications, making comprehension, maintenance, and verification costly and error-prone. While prior work has explored generating specifications from RTL using large language models (LLMs), ensuring that the generated documents faithfully capture design intent remains a major challenge. We present SpecLoop, an agentic framework for RTL-to-specification generation with a formal-verification-driven iterative feedback loop. SpecLoop first generates candidate specifications and then reconstructs RTL from these specifications; it uses formal equivalence checking tools between the reconstructed RTL and the original design to validate functional consistency. When mismatches are detected, counterexamples are fed back to iteratively refine the specifications until equivalence is proven or no further progress can be made. Experiments across multiple LLMs and RTL benchmarks show that incorporating formal verification feedback substantially improves specification correctness and robustness over LLM-only baselines, demonstrating the effectiveness of verification-guided specification generation. △ Less

Submitted 3 March, 2026; originally announced March 2026.

Independence-Number Parameterized Space Complexity for Directed Connectivity Certificate

Authors: Ho-Lin Chen, Tsun Ming Cheung, Peng-Ting Lin, Meng-Tsung Tsai

Abstract: We study the space complexity of computing a sparse subgraph of a directed graph that certifies connectivity in the streaming and distributed models. Formally, for a directed graph $G=(V,A)$ and $k\in \mathbb{N}$, a $k$-node strong connectivity certificate is a subgraph $H=(V,A')\subseteq G$ such that for every pair of distinct nodes $s,t\in V$, the number of pairwise internally node-disjoint path… ▽ More We study the space complexity of computing a sparse subgraph of a directed graph that certifies connectivity in the streaming and distributed models. Formally, for a directed graph $G=(V,A)$ and $k\in \mathbb{N}$, a $k$-node strong connectivity certificate is a subgraph $H=(V,A')\subseteq G$ such that for every pair of distinct nodes $s,t\in V$, the number of pairwise internally node-disjoint paths from $s$ to $t$ in $H$ is at least $k$ or the corresponding number in $G$. In light of the inherent hardness of directed connectivity problems, several prior work focused on restricted graph classes, showing that several problems that are hard in general become efficiently solvable when the input graph is a tournament (i.e., a directed complete graph) (Chakrabarti et al. [SODA 2020]; Baweja, Jia, and Woddruff [ITCS 2022]), or close to a tournament in edit distance (Ghosh and Kuchlous [ESA 2024]). Extending this line of work, our main result shows, at a qualitative level, that the streaming complexity of strong connectivity certificates and related problems is parameterized by independence number, demonstrating a continuum of hardness for directed graph connectivity problems. Quantitatively, for an $n$-node graph with independence number $α$, we give $p$-pass randomized algorithms that compute a $k$-node strong connectivity certificate of size $O(αn)$ using $\tilde{O}(k^{1-1/p}αn^{1+1/p})$ space in the insertion-only model. For the lower bound, we show that even when $k=1$, any $p$-pass streaming algorithm for a 1-node strong connectivity certificate in the insertion-only model requires $Ω(αn/p)$ space. To derive these lower bounds, we introduce the gadget-embedding tournament framework to construct direct-sum-type hard instances with a prescribed independence number, which is applicable to lower-bounding a wide range of directed graph problems. △ Less

Submitted 13 February, 2026; originally announced February 2026.

Efficient Streaming Algorithms for Two-Dimensional Congruence Testing and Geometric Hashing

Authors: Yen-Cheng Chang, Tsun Ming Cheung, Meng-Tsung Tsai, Ting-An Wu

Abstract: The geometric congruence problem is a fundamental building block in many computer vision and image recognition tasks. This problem considers the decision task of whether two point sets are congruent under translation and rotation. A related and more general problem, geometric hashing, considers the task of compactly encoding multiple point sets for efficient congruence queries. Despite its wide ap… ▽ More The geometric congruence problem is a fundamental building block in many computer vision and image recognition tasks. This problem considers the decision task of whether two point sets are congruent under translation and rotation. A related and more general problem, geometric hashing, considers the task of compactly encoding multiple point sets for efficient congruence queries. Despite its wide applications, both problems have received little prior attention in space-aware settings. In this work, we study the two-dimensional congruence testing and geometric hashing problem in the streaming model, where data arrive as a stream and the primary goal is to minimize the space usage. To meaningfully analyze space complexity, we address the underaddressed issue of input precision by working in the finite-precision rational setting: the input point coordinates are rational numbers of the form $p/q$ with $|p|, |q| \le U$. Our result considers a stronger variant of congruence testing called congruence identification, for which we obtain a 3-pass randomized streaming algorithm using $O(\log n(\log U+\log n))$ space. Using the congruence identification algorithm as a building block, we give a 6-pass $O(m\log n (\log n + \log U + \log m))$-space randomized streaming algorithm that outputs a hash function of length $O(\log n+\log U+\log m)$. Our key technical tool for achieving space efficiency is the use of complex moments. While complex moment methods are widely employed as heuristics in object recognition, their effectiveness is often limited by vanishing moment issues (Flusser and Suk [IEEE Trans. Image Process 2006]). We show that, in the rational setting, it suffices to track only $O(\log n)$ complex moments to ensure a non-vanishing moment, thus providing a sound theoretical guarantee for recovering a valid rotation in positive instances. △ Less

Submitted 13 February, 2026; originally announced February 2026.

Space Complexity Dichotomies for Subgraph Finding Problems in the Streaming Model

Authors: Yu-Sheng Shih, Meng-Tsung Tsai, Yen-Chu Tsai, Ying-Sian Wu

Abstract: We study the space complexity of four variants of the standard subgraph finding problem in the streaming model. Specifically, given an $n$-vertex input graph and a fixed-size pattern graph, we consider two settings: undirected simple graphs, denoted by $G$ and $H$, and oriented graphs, denoted by $\vec{G}$ and $\vec{H}$. Depending on the setting, the task is to decide whether $G$ contains $H$ as a… ▽ More We study the space complexity of four variants of the standard subgraph finding problem in the streaming model. Specifically, given an $n$-vertex input graph and a fixed-size pattern graph, we consider two settings: undirected simple graphs, denoted by $G$ and $H$, and oriented graphs, denoted by $\vec{G}$ and $\vec{H}$. Depending on the setting, the task is to decide whether $G$ contains $H$ as a subgraph or as an induced subgraph, or whether $\vec{G}$ contains $\vec{H}$ as a subgraph or as an induced subgraph. Let Sub$(H)$, IndSub$(H)$, Sub$(\vec{H})$, and IndSub$(\vec{H})$ denote these four variants, respectively. An oriented graph is well-oriented if it admits a bipartition in which every arc is oriented from one part to the other, and a vertex is non-well-oriented if both its in-degree and out-degree are non-zero. For each variant, we obtain a complete dichotomy theorem, briefly summarized as follows. (1) Sub$(H)$ can be solved by an $\tilde{O}(1)$-pass $n^{2-Ω(1)}$-space algorithm if and only if $H$ is bipartite. (2) IndSub$(H)$ can be solved by an $\tilde{O}(1)$-pass $n^{2-Ω(1)}$-space algorithm if and only if $H \in \{P_3, P_4, co\mbox{-}P_3\}$. (3) Sub$(\vec{H})$ can be solved by a single-pass $n^{2-Ω(1)}$-space algorithm if and only if every connected component of $\vec H$ is either a well-oriented bipartite graph or a tree containing at most one non-well-oriented vertex. (4) IndSub$(\vec{H})$ can be solved by an $\tilde{O}(1)$-pass $n^{2-Ω(1)}$-space algorithm if and only if the underlying undirected simple graph $H$ is a $co\mbox{-}P_3$. △ Less

Submitted 8 February, 2026; originally announced February 2026.

Determining the Outerthickness of Graphs Is NP-Hard

Authors: Pin-Hsian Lee, Te-Cheng Liu, Meng-Tsung Tsai

Abstract: We give a short, self-contained, and easily verifiable proof that determining the outerthickness of a general graph is NP-hard. This resolves a long-standing open problem on the computational complexity of outerthickness. Moreover, our hardness result applies to a more general covering problem $P_F$, defined as follows. Fix a proper graph class $F$ whose membership is decidable. Given an undirec… ▽ More We give a short, self-contained, and easily verifiable proof that determining the outerthickness of a general graph is NP-hard. This resolves a long-standing open problem on the computational complexity of outerthickness. Moreover, our hardness result applies to a more general covering problem $P_F$, defined as follows. Fix a proper graph class $F$ whose membership is decidable. Given an undirected simple graph $G$ and an integer $k$, the task is to cover the edge set $E(G)$ by at most $k$ subsets $E_1,\ldots,E_k$ such that each subgraph $(V(G),E_i)$ belongs to $F$. Note that if $F$ is monotone (in particular, when $F$ is the class of all outerplanar graphs), any such cover can be converted into an edge partition by deleting overlaps; hence, in this case, covering and partitioning are equivalent. Our result shows that for every proper graph class $F$ whose membership is decidable and that satisfies all of the following conditions: (a) $F$ is closed under topological minors, (b) $F$ is closed under $1$-sums, and (c) $F$ contains a cycle of length $3$, the problem $P_F$ is NP-hard for every fixed integer $k\ge 3$. In particular: For $F$ equal to the class of all outerplanar graphs, our result settles the long-standing open problem on the complexity of determining outerthickness. For $F$ equal to the class of all planar graphs, our result complements Mansfield's NP-hardness result for the thickness, which applies only to the case $k=2$. It is also worth noting that each of the three conditions above is necessary. If $F$ is the class of all eulerian graphs, then cond. (a) fails. If $F$ is the class of all pseudoforests, then cond. (b) fails. If $F$ is the class of all forests, then cond. (c) fails. For each of these three classes $F$, the problem $P_F$ is solvable in polynomial time for every fixed integer $k\ge 3$, showing that none of the three conditions can be dropped. △ Less

Submitted 7 February, 2026; originally announced February 2026.

Time-domain measurement of Auger electron dynamics in xenon atoms after giant resonant photoionization

Authors: Mahmudul Hasan, Jingsong Gao, Hao Liang, Yiming Yuan, Zach Eisenhutt, Ming-Shian Tsai, Ming-Chang Chen, Hans Jakob Wörner, Artem Rudenko, Meng Han

Abstract: Time-resolved measurement of Auger-Meitner (AM) decay [Nature 419, 803 (2002)] marked a milestone in the development of attosecond science. To date, the time constants for the AM decay processes obtained from the time-domain experiments were found to be consistent with the values deduced from conventional energy-domain measurements. One of the main factors limiting the temporal resolution of these… ▽ More Time-resolved measurement of Auger-Meitner (AM) decay [Nature 419, 803 (2002)] marked a milestone in the development of attosecond science. To date, the time constants for the AM decay processes obtained from the time-domain experiments were found to be consistent with the values deduced from conventional energy-domain measurements. One of the main factors limiting the temporal resolution of these studies is the unlocked carrier-envelope-phase (CEP) of the laser pulses used to probe the electronic dynamics triggered by inner-shell photoabsorption. In this work, we report time-resolved inner-shell electron spectroscopy of xenon and krypton using attosecond soft X-ray (atto-SXR) pulses centered at 130 eV in combination with CEP-stabilized few-cycle Yb laser pulses. We observed that the N$_{4,5}$OO Auger electrons from xenon exhibit a clear streaking pattern, but with an unexpected time shift of $\sim$ 1.32 fs relative to the 4$d$ photoelectrons. Furthermore, the energy-integrated yield of streaked Auger electrons from xenon exhibits a pronounced minimum at a pump-probe time delay of 4 fs. Neither of these observations can be explained by current streaking theories and both are inconsistent with lifetimes inferred from energy-domain measurements. The M$_{4,5}$NN Auger electrons from krypton partly overlap in energy with the 3$d$ inner-shell photoelectrons and do not show these anomalous features. This study offers new insights into the inner-shell electron dynamics of heavy atoms in the giant dipole resonance region, laying the groundwork for attosecond soft X-ray spectroscopy of molecular systems containing iodine or bromine atoms. △ Less

Submitted 22 December, 2025; originally announced December 2025.

Comments: 4 figures

Using Vision-Language Models as Proxies for Social Intelligence in Human-Robot Interaction

Authors: Fanjun Bu, Melina Tsai, Audrey Tjokro, Tapomayukh Bhattacharjee, Jorge Ortiz, Wendy Ju

Abstract: Robots operating in everyday environments must often decide when and whether to engage with people, yet such decisions often hinge on subtle nonverbal cues that unfold over time and are difficult to model explicitly. Drawing on a five-day Wizard-of-Oz deployment of a mobile service robot in a university cafe, we analyze how people signal interaction readiness through nonverbal behaviors and how ex… ▽ More Robots operating in everyday environments must often decide when and whether to engage with people, yet such decisions often hinge on subtle nonverbal cues that unfold over time and are difficult to model explicitly. Drawing on a five-day Wizard-of-Oz deployment of a mobile service robot in a university cafe, we analyze how people signal interaction readiness through nonverbal behaviors and how expert wizards use these cues to guide engagement. Motivated by these observations, we propose a two-stage pipeline in which lightweight perceptual detectors (gaze shifts and proxemics) are used to selectively trigger heavier video-based vision-language model (VLM) queries at socially meaningful moments. We evaluate this pipeline on replayed field interactions and compare two prompting strategies. Our findings suggest that selectively using VLMs as proxies for social reasoning enables socially responsive robot behavior, allowing robots to act appropriately by attending to the cues people naturally provide in real-world interactions. △ Less

Submitted 8 December, 2025; originally announced December 2025.

Assessing Large Language Models in Generating RTL Design Specifications

Authors: Hung-Ming Huang, Yu-Hsin Yang, Fu-Chieh Chang, Yun-Chia Hsu, Yin-Yu Lin, Ming-Fang Tsai, Chun-Chih Yang, Pei-Yuan Wu

Abstract: As IC design grows more complex, automating comprehension and documentation of RTL code has become increasingly important. Engineers currently should manually interpret existing RTL code and write specifications, a slow and error-prone process. Although LLMs have been studied for generating RTL from specifications, automated specification generation remains underexplored, largely due to the lack o… ▽ More As IC design grows more complex, automating comprehension and documentation of RTL code has become increasingly important. Engineers currently should manually interpret existing RTL code and write specifications, a slow and error-prone process. Although LLMs have been studied for generating RTL from specifications, automated specification generation remains underexplored, largely due to the lack of reliable evaluation methods. To address this gap, we investigate how prompting strategies affect RTL-to-specification quality and introduce metrics for faithfully evaluating generated specs. We also benchmark open-source and commercial LLMs, providing a foundation for more automated and efficient specification workflows in IC design. △ Less

Submitted 17 November, 2025; originally announced December 2025.

Defending Unauthorized Model Merging via Dual-Stage Weight Protection

Authors: Wei-Jia Chen, Min-Yen Tsai, Cheng-Yi Lee, Chia-Mu Yu

Abstract: The rapid proliferation of pretrained models and open repositories has made model merging a convenient yet risky practice, allowing free-riders to combine fine-tuned models into a new multi-capability model without authorization. Such unauthorized model merging not only violates intellectual property rights but also undermines model ownership and accountability. To address this issue, we present M… ▽ More The rapid proliferation of pretrained models and open repositories has made model merging a convenient yet risky practice, allowing free-riders to combine fine-tuned models into a new multi-capability model without authorization. Such unauthorized model merging not only violates intellectual property rights but also undermines model ownership and accountability. To address this issue, we present MergeGuard, a proactive dual-stage weight protection framework that disrupts merging compatibility while maintaining task fidelity. In the first stage, we redistribute task-relevant information across layers via L2-regularized optimization, ensuring that important gradients are evenly dispersed. In the second stage, we inject structured perturbations to misalign task subspaces, breaking curvature compatibility in the loss landscape. Together, these stages reshape the model's parameter geometry such that merged models collapse into destructive interference while the protected model remains fully functional. Extensive experiments on both vision (ViT-L-14) and language (Llama2, Gemma2, Mistral) models demonstrate that MergeGuard reduces merged model accuracy by up to 90% with less than 1.5% performance loss on the protected model. △ Less

Submitted 12 March, 2026; v1 submitted 14 November, 2025; originally announced November 2025.

Comments: Accepted at CVPR 2026, updated

MuFFIN: Multifaceted Pronunciation Feedback Model with Interactive Hierarchical Neural Modeling

Authors: Bi-Cheng Yan, Ming-Kang Tsai, Berlin Chen

Abstract: Computer-assisted pronunciation training (CAPT) manages to facilitate second-language (L2) learners to practice pronunciation skills by offering timely and instructive feedback. To examine pronunciation proficiency from multiple facets, existing methods for CAPT broadly fall into two categories: mispronunciation detection and diagnosis (MDD) as well as automatic pronunciation assessment (APA). The… ▽ More Computer-assisted pronunciation training (CAPT) manages to facilitate second-language (L2) learners to practice pronunciation skills by offering timely and instructive feedback. To examine pronunciation proficiency from multiple facets, existing methods for CAPT broadly fall into two categories: mispronunciation detection and diagnosis (MDD) as well as automatic pronunciation assessment (APA). The former aims to pinpoint phonetic pronunciation errors and provide diagnostic feedback, while the latter seeks instead to quantify pronunciation proficiency pertaining to various aspects. Despite the natural complementarity between MDD and APA, researchers and practitioners, however, often treat them as independent tasks with disparate modeling paradigms. In light of this, we in this paper first introduce MuFFIN, a Multi-Faceted pronunciation Feedback model with an Interactive hierarchical Neural architecture, to jointly address the tasks of MDD and APA. To better capture the nuanced distinctions between phonemes in the feature space, a novel phoneme-contrastive ordinal regularization mechanism is then put forward to optimize the proposed model to generate more phoneme-discriminative features while factoring in the ordinality of the aspect scores. In addition, to address the intricate data imbalance problem in MDD, we design a simple yet effective training objective, which is specifically tailored to perturb the outputs of a phoneme classifier with the phoneme-specific variations, so as to better render the distribution of predicted phonemes meanwhile considering their mispronunciation characteristics. A series of experiments conducted on the Speechocean762 benchmark dataset demonstrates the efficacy of our method in relation to several cutting-edge baselines, showing state-of-the-art performance on both the APA and MDD tasks. △ Less

Submitted 7 October, 2025; v1 submitted 6 October, 2025; originally announced October 2025.

Comments: Accepted and to appear in IEEE/ACM Transactions on Audio, Speech, and Language Processing

Multiplicative trace and spectrum preservers on stochastic matrices

Authors: Ming-Cheng Tsai, Huajun Huang

Abstract: We characterize maps $φ_i: \mathcal{S} \to \mathcal{S}$, $i=1, \ldots, m$ and $m\ge 1$, that have the multiplicative spectrum or trace preserving property: \begin{eqnarray*} \textrm{spec} (φ_1(A_1)\cdots φ_m(A_m)) &=& \textrm{spec} (A_1\cdots A_m),\quad\text{or}\quad \textrm{tr} (φ_1(A_1)\cdots φ_m(A_m)) &=& \textrm{tr} (A_1\cdots A_m), \end{eqnarray*} where $\mathcal{S}$ is the set of… ▽ More We characterize maps $φ_i: \mathcal{S} \to \mathcal{S}$, $i=1, \ldots, m$ and $m\ge 1$, that have the multiplicative spectrum or trace preserving property: \begin{eqnarray*} \textrm{spec} (φ_1(A_1)\cdots φ_m(A_m)) &=& \textrm{spec} (A_1\cdots A_m),\quad\text{or}\quad \textrm{tr} (φ_1(A_1)\cdots φ_m(A_m)) &=& \textrm{tr} (A_1\cdots A_m), \end{eqnarray*} where $\mathcal{S}$ is the set of $n\times n$ doubly stochastic, row stochastic, or column stochastic matrices, or the space spanned by one of these sets. Linearity is assumed when $m=1$. We show that every stochastic matrix contains a real doubly stochastic component that carries the spectral information. In consequence, the multiplicative spectrum or trace preservers on these sets $ \mathcal{S} $ are linked to the corresponding preservers on the space of doubly stochastic matrices. Moreover, when $m\ge 3$, multiplicative trace preservers always coincide with multiplicative spectrum preservers. △ Less

Submitted 25 September, 2025; originally announced September 2025.

Comments: 31 pages

MSC Class: Primary 15A86; Secondary 47B49; 15A15; 15A18

Model selection meets clinical semantics: Optimizing ICD-10-CM prediction via LLM-as-Judge evaluation, redundancy-aware sampling, and section-aware fine-tuning

Authors: Hong-Jie Dai, Zheng-Hao Li, An-Tai Lu, Bo-Tsz Shain, Ming-Ta Li, Tatheer Hussain Mir, Kuang-Te Wang, Min-I Su, Pei-Kang Liu, Ming-Ju Tsai

Abstract: Accurate International Classification of Diseases (ICD) coding is critical for clinical documentation, billing, and healthcare analytics, yet it remains a labour-intensive and error-prone task. Although large language models (LLMs) show promise in automating ICD coding, their challenges in base model selection, input contextualization, and training data redundancy limit their effectiveness. We pro… ▽ More Accurate International Classification of Diseases (ICD) coding is critical for clinical documentation, billing, and healthcare analytics, yet it remains a labour-intensive and error-prone task. Although large language models (LLMs) show promise in automating ICD coding, their challenges in base model selection, input contextualization, and training data redundancy limit their effectiveness. We propose a modular framework for ICD-10 Clinical Modification (ICD-10-CM) code prediction that addresses these challenges through principled model selection, redundancy-aware data sampling, and structured input design. The framework integrates an LLM-as-judge evaluation protocol with Plackett-Luce aggregation to assess and rank open-source LLMs based on their intrinsic comprehension of ICD-10-CM code definitions. We introduced embedding-based similarity measures, a redundancy-aware sampling strategy to remove semantically duplicated discharge summaries. We leverage structured discharge summaries from Taiwanese hospitals to evaluate contextual effects and examine section-wise content inclusion under universal and section-specific modelling paradigms. Experiments across two institutional datasets demonstrate that the selected base model after fine-tuning consistently outperforms baseline LLMs in internal and external evaluations. Incorporating more clinical sections consistently improves prediction performance. This study uses open-source LLMs to establish a practical and principled approach to ICD-10-CM code prediction. The proposed framework provides a scalable, institution-ready solution for real-world deployment of automated medical coding systems by combining informed model selection, efficient data refinement, and context-aware prompting. △ Less

Submitted 23 September, 2025; originally announced September 2025.

Comments: 28 Pages, 4 Figures, 2 Tables

ACM Class: I.2.6; I.2.7; J.3

CFDA & CLIP at TREC iKAT 2025: Enhancing Personalized Conversational Search via Query Reformulation and Rank Fusion

Authors: Yu-Cheng Chang, Guan-Wei Yeo, Quah Eugene, Fan-Jie Shih, Yuan-Ching Kuo, Tsung-En Yu, Hung-Chun Hsu, Ming-Feng Tsai, Chuan-Ju Wang

Abstract: The 2025 TREC Interactive Knowledge Assistance Track (iKAT) featured both interactive and offline submission tasks. The former requires systems to operate under real-time constraints, making robustness and efficiency as important as accuracy, while the latter enables controlled evaluation of passage ranking and response generation with pre-defined datasets. To address this, we explored query rewri… ▽ More The 2025 TREC Interactive Knowledge Assistance Track (iKAT) featured both interactive and offline submission tasks. The former requires systems to operate under real-time constraints, making robustness and efficiency as important as accuracy, while the latter enables controlled evaluation of passage ranking and response generation with pre-defined datasets. To address this, we explored query rewriting and retrieval fusion as core strategies. We built our pipelines around Best-of-$N$ selection and Reciprocal Rank Fusion (RRF) strategies to handle different submission tasks. Results show that reranking and fusion improve robustness while revealing trade-offs between effectiveness and efficiency across both tasks. △ Less

Submitted 19 September, 2025; originally announced September 2025.

Strong-Field Photoelectron Interferometry with Near-Single-Cycle Yb Lasers

Authors: Mahmudul Hasan, Phi-Hung Tran, Jingsong Gao, Van-Hung Hoang, Ming-Shian Tsai, Ming-Chang Chen, Uwe Thumm, Charles Lewis Cocke, Chii-Dong Lin, Anh-Thu Le, and Meng Han

Abstract: The concept of using photoelectron interferometry in short laser fields to probe electron dynamics and target structures was introduced more than two decades ago. However, the quality of experimental data has remained insufficient for quantitative analysis, largely due to the instability of few-cycle Ti:Sa laser pulses, the current workhorse of short pulses. Here, we report the first systematic st… ▽ More The concept of using photoelectron interferometry in short laser fields to probe electron dynamics and target structures was introduced more than two decades ago. However, the quality of experimental data has remained insufficient for quantitative analysis, largely due to the instability of few-cycle Ti:Sa laser pulses, the current workhorse of short pulses. Here, we report the first systematic strong-field ionization experiments performed with industrial-grade, carrier-envelope-phase (CEP) stabilized, near-single-cycle Yb lasers. By measuring photoelectron momentum distributions in the direct-ionization regime, we show that single-cycle cosine-shaped pulses can separate and enhance both spider-leg and fishbone holographic structures. The spider-leg structure enables extraction of the electron scattering phase from the Ar atomic potential-information typically accessible only through attosecond metrology, while the fishbone structure reveals the orbital-parity contrast between Ar atoms and nitrogen molecules. Our measurements are quantitatively reproduced by both semiclassical Herman-Kluk-propagator and \textit{ab initio} simulations, paving the way for precision studies of electron-molecule scattering with widely accessible industrial-grade lasers. △ Less

Submitted 2 September, 2025; v1 submitted 1 September, 2025; originally announced September 2025.

Comments: 4 figures

Journal ref: Phys. Rev. Lett. 135, 263001 (2025)

Test-Time Scaling Strategies for Generative Retrieval in Multimodal Conversational Recommendations

Authors: Hung-Chun Hsu, Yuan-Ching Kuo, Chao-Han Huck Yang, Szu-Wei Fu, Hanrong Ye, Hongxu Yin, Yu-Chiang Frank Wang, Ming-Feng Tsai, Chuan-Ju Wang

Abstract: The rapid evolution of e-commerce has exposed the limitations of traditional product retrieval systems in managing complex, multi-turn user interactions. Recent advances in multimodal generative retrieval -- particularly those leveraging multimodal large language models (MLLMs) as retrievers -- have shown promise. However, most existing methods are tailored to single-turn scenarios and struggle to… ▽ More The rapid evolution of e-commerce has exposed the limitations of traditional product retrieval systems in managing complex, multi-turn user interactions. Recent advances in multimodal generative retrieval -- particularly those leveraging multimodal large language models (MLLMs) as retrievers -- have shown promise. However, most existing methods are tailored to single-turn scenarios and struggle to model the evolving intent and iterative nature of multi-turn dialogues when applied naively. Concurrently, test-time scaling has emerged as a powerful paradigm for improving large language model (LLM) performance through iterative inference-time refinement. Yet, its effectiveness typically relies on two conditions: (1) a well-defined problem space (e.g., mathematical reasoning), and (2) the model's ability to self-correct -- conditions that are rarely met in conversational product search. In this setting, user queries are often ambiguous and evolving, and MLLMs alone have difficulty grounding responses in a fixed product corpus. Motivated by these challenges, we propose a novel framework that introduces test-time scaling into conversational multimodal product retrieval. Our approach builds on a generative retriever, further augmented with a test-time reranking (TTR) mechanism that improves retrieval accuracy and better aligns results with evolving user intent throughout the dialogue. Experiments across multiple benchmarks show consistent improvements, with average gains of 14.5 points in MRR and 10.6 points in nDCG@1. △ Less

Submitted 25 August, 2025; originally announced August 2025.

Bright 25-attosecond light pulses reach the one atomic unit of time

Authors: Jingsong Gao, Mahmudul Hasan, Hao Liang, Ming-Shian Tsai, Yiming Yuan, Zach Eisenhutt, Christoph H. Keitel, Chii-Dong Lin, Yunquan Liu, Ming-Chang Chen, Meng Han

Abstract: Generating ever-shorter and brighter light pulses has long been a central pursuit in ultrafast science, as it benchmarks our ability to create and manipulate the coherence on the intrinsic timescale of sub-atomic electron motion. The current state-of-the-art in attosecond pulse generation reaches durations of 40-50 attoseconds (1 as = $10^{-18}$ seconds), produced via high-order harmonic generatio… ▽ More Generating ever-shorter and brighter light pulses has long been a central pursuit in ultrafast science, as it benchmarks our ability to create and manipulate the coherence on the intrinsic timescale of sub-atomic electron motion. The current state-of-the-art in attosecond pulse generation reaches durations of 40-50 attoseconds (1 as = $10^{-18}$ seconds), produced via high-order harmonic generation (HHG) driven by secondary mid-infrared light sources. However, these sources often suffer from low stability and poor HHG conversion efficiency. In this work, we demonstrate the generation of 25$\pm$2 attosecond light pulses, a new world record for the shortest light pulse, driven by a post-compressed, industrial-grade Yb-based laser system. The resulting high-harmonic spectrum spans photon energies from 50 eV to 320 eV, covering the carbon K-edge, with a calibrated photon flux exceeding $10^{12}$ photons per second, approximately three orders of magnitude higher than previous studies. The pulse duration was characterized using an angle-resolved photoelectron streaking camera on helium atoms and systematically optimized through the use of dielectric filters of varying thicknesses to compensate the attochirp. Our study reaches the threshold of one atomic unit of time (24.2 attoseconds), the boundary between atomic and ionic physics, opening the door to resolving exciting ionic quantum dynamics with tabletop lasers. △ Less

Submitted 20 August, 2025; originally announced August 2025.

Automating Expert-Level Medical Reasoning Evaluation of Large Language Models

Authors: Shuang Zhou, Wenya Xie, Jiaxi Li, Zaifu Zhan, Meijia Song, Han Yang, Cheyenna Espinoza, Lindsay Welton, Xinnie Mai, Yanwei Jin, Zidu Xu, Yuen-Hei Chung, Yiyun Xing, Meng-Han Tsai, Emma Schaffer, Yucheng Shi, Ninghao Liu, Zirui Liu, Rui Zhang

Abstract: As large language models (LLMs) become increasingly integrated into clinical decision-making, ensuring transparent and trustworthy reasoning is essential. However, existing evaluation strategies of LLMs' medical reasoning capability either suffer from unsatisfactory assessment or poor scalability, and a rigorous benchmark remains lacking. To address this, we introduce MedThink-Bench, a benchmark d… ▽ More As large language models (LLMs) become increasingly integrated into clinical decision-making, ensuring transparent and trustworthy reasoning is essential. However, existing evaluation strategies of LLMs' medical reasoning capability either suffer from unsatisfactory assessment or poor scalability, and a rigorous benchmark remains lacking. To address this, we introduce MedThink-Bench, a benchmark designed for rigorous, explainable, and scalable assessment of LLMs' medical reasoning. MedThink-Bench comprises 500 challenging questions across ten medical domains, each annotated with expert-crafted step-by-step rationales. Building on this, we propose LLM-w-Ref, a novel evaluation framework that leverages fine-grained rationales and LLM-as-a-Judge mechanisms to assess intermediate reasoning with expert-level fidelity while maintaining scalability. Experiments show that LLM-w-Ref exhibits a strong positive correlation with expert judgments. Benchmarking twelve state-of-the-art LLMs, we find that smaller models (e.g., MedGemma-27B) can surpass larger proprietary counterparts (e.g., OpenAI-o3). Overall, MedThink-Bench offers a foundational tool for evaluating LLMs' medical reasoning, advancing their safe and responsible deployment in clinical practice. △ Less

Submitted 10 July, 2025; originally announced July 2025.

Comments: 22 pages,6 figures

Computing Diverse and Nice Triangulations

Authors: Waldo Gálvez, Mayank Goswami, Arturo Merino, GiBeom Park, Meng-Tsung Tsai

Abstract: We initiate the study of computing diverse triangulations to a given polygon. Given a simple $n$-gon $P$, an integer $ k \geq 2 $, a quality measure $σ$ on the set of triangulations of $P$ and a factor $ α\geq 1 $, we formulate the Diverse and Nice Triangulations (DNT) problem that asks to compute $k$ \emph{distinct} triangulations $T_1,\dots,T_k$ of $P$ such that a) their diversity,… ▽ More We initiate the study of computing diverse triangulations to a given polygon. Given a simple $n$-gon $P$, an integer $ k \geq 2 $, a quality measure $σ$ on the set of triangulations of $P$ and a factor $ α\geq 1 $, we formulate the Diverse and Nice Triangulations (DNT) problem that asks to compute $k$ \emph{distinct} triangulations $T_1,\dots,T_k$ of $P$ such that a) their diversity, $\sum_{i < j} d(T_i,T_j) $, is as large as possible \emph{and} b) they are nice, i.e., $σ(T_i) \leq ασ^* $ for all $1\leq i \leq k$. Here, $d$ denotes the symmetric difference of edge sets of two triangulations, and $σ^*$ denotes the best quality of triangulations of $P$, e.g., the minimum Euclidean length. As our main result, we provide a $\mathrm{poly}(n,k)$-time approximation algorithm for the DNT problem that returns a collection of $k$ distinct triangulations whose diversity is at least $1 - Θ(1/k)$ of the optimal, and each triangulation satisfies the quality constraint. This is accomplished by studying \emph{bi-criteria triangulations} (BCT), which are triangulations that simultaneously optimize two criteria, a topic of independent interest. We complement our approximation algorithms by showing that the DNT problem and the BCT problem are NP-hard. Finally, for the version where diversity is defined as $\min_{i < j} d(T_i,T_j) $, we show a reduction from the problem of computing optimal Hamming codes, and provide an $n^{O(k)}$-time $\tfrac12$-approximation algorithm. This improves over the naive ${C_{n-2} \choose k} \approx 2^{O(nk)}$ time bound for enumerating all $k$-tuples among the triangulations of a simple $n$-gon, where $C_n$ denotes the $n$-th Catalan number. △ Less

Submitted 10 June, 2025; v1 submitted 2 June, 2025; originally announced June 2025.

texTENG: Fabricating Wearable Textile-Based Triboelectric Nanogenerators

Authors: Ritik Batra, Narjes Pourjafarian, Samantha Chang, Margaret Tsai, Jacob Revelo, Cindy Hsin-Liu Kao

Abstract: Recently, there has been a surge of interest in sustainable energy sources, particularly for wearable computing. Triboelectric nanogenerators (TENGs) have shown promise in converting human motion into electric power. Textile-based TENGs, valued for their flexibility and breathability, offer an ideal form factor for wearables. However, uptake in maker communities has been slow due to commercially u… ▽ More Recently, there has been a surge of interest in sustainable energy sources, particularly for wearable computing. Triboelectric nanogenerators (TENGs) have shown promise in converting human motion into electric power. Textile-based TENGs, valued for their flexibility and breathability, offer an ideal form factor for wearables. However, uptake in maker communities has been slow due to commercially unavailable materials, complex fabrication processes, and structures incompatible with human motion. This paper introduces texTENG, a textile-based framework simplifying the fabrication of power harvesting and self-powered sensing applications. By leveraging accessible materials and familiar tools, texTENG bridges the gap between advanced TENG research and wearable applications. We explore a design menu for creating multidimensional TENG structures using braiding, weaving, and knitting. Technical evaluations and example applications highlight the performance and feasibility of these designs, offering DIY-friendly pathways for fabricating textile-based TENGs and promoting sustainable prototyping practices within the HCI and maker communities. △ Less

Submitted 16 March, 2025; originally announced March 2025.

Comments: 11 pages

A Framework for the Design of Efficient Diversification Algorithms to NP-Hard Problems

Authors: Waldo Gálvez, Mayank Goswami, Arturo Merino, GiBeom Park, Meng-Tsung Tsai, Victor Verdugo

Abstract: There has been considerable recent interest in computing a diverse collection of solutions to a given optimization problem, both in the AI and theory communities. Given a classical optimization problem $Π$ (e.g., spanning tree, minimum cuts, maximum matching, minimum vertex cover) with input size $n$ and an integer $k\geq 1$, the goal is to generate a collection of $k$ maximally diverse solutions… ▽ More There has been considerable recent interest in computing a diverse collection of solutions to a given optimization problem, both in the AI and theory communities. Given a classical optimization problem $Π$ (e.g., spanning tree, minimum cuts, maximum matching, minimum vertex cover) with input size $n$ and an integer $k\geq 1$, the goal is to generate a collection of $k$ maximally diverse solutions to $Π$. This diverse-X paradigm not only allows the user to generate very different solutions, but also helps make systems more secure and robust by handling uncertainty, and achieve energy efficiency. For problems $Π$ in P (such as spanning tree and minimum cut), there are efficient $\text{poly}(n,k)$ approximation algorithms available for the diverse variants [Hanaka et al. AAAI 2021, 2022, 2023, Gao et al. LATIN 2022, de Berg et al. ISAAC 2023]. In contrast, only FPT algorithms are known for NP-hard problems such as vertex covers and independent sets [Baste et al. IJCAI 2020, Eiben et al. SODA 2024, Misra et al. ISAAC 2024, Austrin et al. ICALP 2025], but in the worst case, these algorithms run in time $\exp((kn)^c)$ for some $c>0$. In this work, we address this gap and give $\text{poly}(n,k)$ or $f(k)\text{poly}(n)$ time approximation algorithms for diversification variants of several NP-hard problems such as knapsack, maximum weight independent sets (MWIS) and minimum vertex covers in planar graphs, geometric (rectangle) knapsack, enclosing points by polygon, and MWIS in unit-disk-graphs of points in convex position. Our results are achieved by developing a general framework and applying it to problems with textbook dynamic-programming algorithms to find one solution. △ Less

Submitted 10 June, 2025; v1 submitted 21 January, 2025; originally announced January 2025.

ACM Class: F.2.2

Design, Construction, and Testing of the APOLLO ATCA Blades for Use at the HL-LHC

Authors: Alp Akpinar, Aymeric Blaizot, Serhii Cholak, Gianfranco de Castro, Zeynep Demiragli, Alec Duquette, Jonathan Richard Fulcher, Dan Gastler, Kristian Hahn, Eric Shearer Hazen, Si Hyun Jeon, Peace Kotamnives, Alexander Madorsky, David Monk, Sheena Noorudhin, Michael Oshiro, James Rohlf, Charles Ralph Strohman, Emily Minyun Tsai, Peter Wittich, Siqi Yuan, Rui Zou

Abstract: The Apollo Advanced Telecommunications Computing Architecture (ATCA) platform is an open-source design consisting of a generic "Service Module" (SM) and a customizable "Command Module" (CM), allowing for cost-effective use in applications such as the readout of the inner tracker and the Level-1 track trigger for the CMS Phase-II upgrade at the HL-LHC. The SM integrates an intelligent IPMC, robust… ▽ More The Apollo Advanced Telecommunications Computing Architecture (ATCA) platform is an open-source design consisting of a generic "Service Module" (SM) and a customizable "Command Module" (CM), allowing for cost-effective use in applications such as the readout of the inner tracker and the Level-1 track trigger for the CMS Phase-II upgrade at the HL-LHC. The SM integrates an intelligent IPMC, robust power entry and conditioning systems, a powerful system-on-module computer, and flexible clock and communication infrastructure. The CM is designed around two Xilinx Ultrascale+ FPGAs and high-density, high-bandwidth optical transceivers capable of 25 Gb/s. Crates of Apollo blades are currently being tested at Boston University, Cornell University, and CERN. △ Less

Submitted 21 March, 2025; v1 submitted 7 January, 2025; originally announced January 2025.

Comments: 8 pages, 4 figures, TWEPP2024 conference proceedings

Filamentation-Assisted Isolated Attosecond Pulse Generation

Authors: Yu-En Chien, Marina Fernández-Galán, Ming-Shian Tsai, An-Yuan Liang, Enrique Conejero-Jarque, Javier Serrano, Julio San Román, Carlos Hernández-García, Ming-Chang Chen

Abstract: Isolated attosecond pulses (IAPs) generated by few-cycle femtosecond lasers are essential for capturing ultrafast dynamics in atoms, molecules, and solids. Nonetheless, the advancement of attosecond science critically depends on achieving stable, high-temporal-contrast IAPs. Our study reveals a universal scenario in which self-compression of the infrared driver in high harmonic generation in exten… ▽ More Isolated attosecond pulses (IAPs) generated by few-cycle femtosecond lasers are essential for capturing ultrafast dynamics in atoms, molecules, and solids. Nonetheless, the advancement of attosecond science critically depends on achieving stable, high-temporal-contrast IAPs. Our study reveals a universal scenario in which self-compression of the infrared driver in high harmonic generation in extended gas media leads to high-contrast high-frequency IAP generation. Our experimental and theoretical results reveal that filamentation in a semi-infinite gas cell not only shapes the infrared driving pulse spatially and temporally, but also creates a stable propagation region where high harmonic generation is phase-matched, leading to the production of bright IAPs. In an argon-filled gas cell, filamentation notably reduces the pulse duration of Yb-based 1030 nm pulses from 4.7 fs to 3.5 fs, while simultaneously generating high-contrast 200-attosecond IAPs at 70 eV. We demonstrate the universality of filamentation-assisted IAP generation, showing that post-compressed Yb-based laser filaments in neon and helium yield even shorter IAPs: 69-attoseconds at 100 eV, and 65-attoseconds IAPs at 135 eV, respectively. This spatiotemporal reshaping of few-cycle pulses through filamentation possesses immediate impacts on both post-compression techniques and attosecond-based technologies. △ Less

Submitted 9 December, 2024; originally announced December 2024.

Towards a unified theory for testing statistical hypothesis: Multinormal mean with nuisance covariance matrix

Authors: Ming-Tien Tsai

Abstract: Under a multinormal distribution with an arbitrary unknown covariance matrix, the main purpose of this paper is to propose a framework to achieve the goal of reconciliation of Bayesian, frequentist, and Fisher's reporting $p$-values, Neyman-Pearson's optimal theory and Wald's decision theory for the problems of testing mean against restricted alternatives (closed convex cones). To proceed, the tes… ▽ More Under a multinormal distribution with an arbitrary unknown covariance matrix, the main purpose of this paper is to propose a framework to achieve the goal of reconciliation of Bayesian, frequentist, and Fisher's reporting $p$-values, Neyman-Pearson's optimal theory and Wald's decision theory for the problems of testing mean against restricted alternatives (closed convex cones). To proceed, the tests constructed via the likelihood ratio (LR) and the union-intersection (UI) principles are studied. For the problems of testing against restricted alternatives, first, we show that the LRT and the UIT are not the proper Bayes tests, however, they are shown to be the integrated LRT and the integrated UIT, respectively. For the problem of testing against the positive orthant space alternative, both the null distributions of the LRT and the UIT depend on the unknown nuisance covariance matrix. Hence we have difficulty adopting Fisher's approach to reporting $p$-values. On the other hand, according to the definition of the level of significance, both the LRT and the UIT are shown to be power-dominated by the corresponding LRT and UIT for testing against the half-space alternative, respectively. Hence, both the LRT and the UIT are $α$-inadmissible, these results are against the common statistical sense. Neither Fisher's approach of reporting $p$-values alone nor Neyman-Pearson's optimal theory for power function alone is a satisfactory criterion for evaluating the performance of tests. Wald's decision theory via $d$-admissibility may shed light on resolving these challenging issues of imposing the balance between type 1 error and power. △ Less

Submitted 7 December, 2024; v1 submitted 19 November, 2024; originally announced November 2024.

Comments: arXiv admin note: substantial text overlap with arXiv:1710.06573

Hearing carrier-envelope offset frequency and phase in air with a microphone

Authors: Meng Han, Ming-Chang Chen, Ming-Shian Tsai, Hao Liang

Abstract: Attosecond science and frequency metrology rely on the precise measurement and control of the laser pulse waveform, a feat traditionally achieved using optoelectronic techniques. In this study, we conducted a laser-induced acoustic experiment in air ionized by carrier-envelope phase (CEP)-stabilized sub-4 femtosecond pulses. Our results reveal that the acoustic signal exhibits CEP dependence in fe… ▽ More Attosecond science and frequency metrology rely on the precise measurement and control of the laser pulse waveform, a feat traditionally achieved using optoelectronic techniques. In this study, we conducted a laser-induced acoustic experiment in air ionized by carrier-envelope phase (CEP)-stabilized sub-4 femtosecond pulses. Our results reveal that the acoustic signal exhibits CEP dependence in few-cycle pulses, primarily through amplitude modulation from laser-driven ionization. This novel optoacoustic phenomenon enables not only the measurement of the carrier-envelope offset frequency but also the direct characterization of the waveform of optical pulses through a microphone. Our study highlights the potential of laser-induced acoustic waves for advancing frequency metrology and ultrafast science. △ Less

Submitted 29 March, 2025; v1 submitted 12 November, 2024; originally announced November 2024.

Comments: 5 figures

ADIOS: Antibody Development via Opponent Shaping

Authors: Sebastian Towers, Aleksandra Kalisz, Philippe A. Robert, Alicia Higueruelo, Francesca Vianello, Ming-Han Chloe Tsai, Harrison Steel, Jakob N. Foerster

Abstract: Anti-viral therapies are typically designed to target only the current strains of a virus, a myopic response. However, therapy-induced selective pressures drive the emergence of new viral strains, against which the original myopic therapies are no longer effective. This evolutionary response presents an opportunity: our therapies could both defend against and actively influence viral evolution. Th… ▽ More Anti-viral therapies are typically designed to target only the current strains of a virus, a myopic response. However, therapy-induced selective pressures drive the emergence of new viral strains, against which the original myopic therapies are no longer effective. This evolutionary response presents an opportunity: our therapies could both defend against and actively influence viral evolution. This motivates our method ADIOS: Antibody Development vIa Opponent Shaping. ADIOS is a meta-learning framework where the process of antibody therapy design, the outer loop, accounts for the virus's adaptive response, the inner loop. With ADIOS, antibodies are not only robust against potential future variants, they also influence, i.e., shape, which future variants emerge. In line with the opponent shaping literature, we refer to our optimised antibodies as shapers. To demonstrate the value of ADIOS, we build a viral evolution simulator using the Absolut! framework, in which shapers successfully target both current and future viral variants, outperforming myopic antibodies. Furthermore, we show that shapers modify the distribution over viral evolutionary trajectories to result in weaker variants. We believe that our ADIOS paradigm will facilitate the discovery of long-lived vaccines and antibody therapies while also generalising to other domains. Specifically, domains such as antimicrobial resistance, cancer treatment, and others with evolutionarily adaptive opponents. Our code is available at https://github.com/olakalisz/adios. △ Less

Submitted 6 June, 2025; v1 submitted 16 September, 2024; originally announced September 2024.

Comments: Accepted at ICML 2025

MSC Class: 92-08 ACM Class: I.2.1; J.3

Journal ref: Proceedings of the 42nd International Conference on Machine Learning (ICML 2025), PMLR 267

Zero-Shot Text-to-Speech as Golden Speech Generator: A Systematic Framework and its Applicability in Automatic Pronunciation Assessment

Authors: Tien-Hong Lo, Meng-Ting Tsai, Yao-Ting Sung, Berlin Chen

Abstract: Second language (L2) learners can improve their pronunciation by imitating golden speech, especially when the speech that aligns with their respective speech characteristics. This study explores the hypothesis that learner-specific golden speech generated with zero-shot text-to-speech (ZS-TTS) techniques can be harnessed as an effective metric for measuring the pronunciation proficiency of L2 lear… ▽ More Second language (L2) learners can improve their pronunciation by imitating golden speech, especially when the speech that aligns with their respective speech characteristics. This study explores the hypothesis that learner-specific golden speech generated with zero-shot text-to-speech (ZS-TTS) techniques can be harnessed as an effective metric for measuring the pronunciation proficiency of L2 learners. Building on this exploration, the contributions of this study are at least two-fold: 1) design and development of a systematic framework for assessing the ability of a synthesis model to generate golden speech, and 2) in-depth investigations of the effectiveness of using golden speech in automatic pronunciation assessment (APA). Comprehensive experiments conducted on the L2-ARCTIC and Speechocean762 benchmark datasets suggest that our proposed modeling can yield significant performance improvements with respect to various assessment metrics in relation to some prior arts. To our knowledge, this study is the first to explore the role of golden speech in both ZS-TTS and APA, offering a promising regime for computer-assisted pronunciation training (CAPT). △ Less

Submitted 26 July, 2025; v1 submitted 11 September, 2024; originally announced September 2024.

Comments: SLaTE 2025

Linear maps on matrices preserving parallel pairs

Authors: Chi-Kwong Li, Ming-Cheng Tsai, Ya-Shu Wang, Ngai-Ching Wong

Abstract: Two (real or complex) $m\times n$ matrices $A$ and $B$ are said to be parallel (resp. triangle equality attaining, or TEA in short) with respect to the spectral norm $\|\cdot\|$ if $\|A+ μB\| = \|A\| + \|B\|$ for some scalar $μ$ with $|μ|=1$ (resp. $μ=1$). We study linear maps $T$ on $m\times n$ matrices preserving parallel (resp. TEA) pairs, i.e., $T(A)$ and $T(B)$ are parallel (resp. TEA) whenev… ▽ More Two (real or complex) $m\times n$ matrices $A$ and $B$ are said to be parallel (resp. triangle equality attaining, or TEA in short) with respect to the spectral norm $\|\cdot\|$ if $\|A+ μB\| = \|A\| + \|B\|$ for some scalar $μ$ with $|μ|=1$ (resp. $μ=1$). We study linear maps $T$ on $m\times n$ matrices preserving parallel (resp. TEA) pairs, i.e., $T(A)$ and $T(B)$ are parallel (resp. TEA) whenever $A$ and $B$ are parallel (resp. TEA). It is shown that when $m,n \ge 2$ and $(m,n) \ne (2,2)$, a nonzero linear map $T$ preserving TEA pairs if and only if it is a positive multiple of a linear isometry, namely, $T$ has the form $$(1) \quad A \mapsto γUAV \quad \quad \text{or} \quad \quad (2) \quad A \mapsto γUA^{t} V \quad (\text{in this case}, m = n),$$ for a positive number $γ$, and unitary (or real orthogonal) matrices $U$ and $V$ of appropriate sizes. Linear maps preserving parallel pairs are those carrying form (1), (2), or the form $$ (3) \ A \mapsto f(A) Z$$ for a linear functional $f$ and a fixed matrix $Z$. The case when $(m,n) = (2,2)$ is more complicated. There are linear maps of $2\times 2$ matrices preserving parallel pairs or TEA pairs neither of the form (1), (2) nor (3) above. Complete characterization of such maps is given with some intricate computation and techniques in matrix groups. △ Less

Submitted 27 July, 2024; originally announced August 2024.

MSC Class: 15A86; 15A60

Linear maps preserving $\ell_p$-norm parallel vectors

Authors: Chi-Kwong Li, Ming-Cheng Tsai, Ya-Shu Wang, Ngai-Ching Wong

Abstract: Two vectors $x, y$ in a normed vector space are parallel if there is a scalar $μ$ with $|μ| = 1$ such that $\|x+μy\| = \|x\| + \|y\|$; they form a triangle equality attaining (TEA) pair if $\|x+y\| = \|x\| + \|y\|$. In this paper, we characterize linear maps on $F^n=R^n$ or $C^n$, equipped with the $\ell_p$-norm for $p \in [1, \infty]$, preserving parallel pairs or preserving TEA pairs. Indeed, an… ▽ More Two vectors $x, y$ in a normed vector space are parallel if there is a scalar $μ$ with $|μ| = 1$ such that $\|x+μy\| = \|x\| + \|y\|$; they form a triangle equality attaining (TEA) pair if $\|x+y\| = \|x\| + \|y\|$. In this paper, we characterize linear maps on $F^n=R^n$ or $C^n$, equipped with the $\ell_p$-norm for $p \in [1, \infty]$, preserving parallel pairs or preserving TEA pairs. Indeed, any linear map will preserve parallel pairs and TEA pairs when $1< p <\infty$. For the $\ell_1$-norm, TEA preservers form a semigroup of matrices in which each row has at most one nonzero entries; adding rank one matrices to this semigroup will be the semigroup of parallel preserves. For the $\ell_\infty$-norm, a nonzero TEA preserver, or a parallel preserver of rank greater than one, is always a multiple of an $\ell_\infty$-norm isometry, except when $F^n = R^2$. We also have a characterization for the exceptional case. The results are extended to linear maps of the infinite dimensional spaces $\ell_1(Λ)$, $c_0(Λ)$ and $\ell_\infty(Λ)$. △ Less

Submitted 27 July, 2024; originally announced July 2024.

MSC Class: 15A86; 15A60

On the consistent estimators of the population covariance matrix and its reparameterizations

Authors: Ming-Tien Tsai, Chia-Hsian Tsai

Abstract: For the high-dimensional covariance estimation problem, when $\lim_{n\to \infty}p/n=c \in (0,1)$ the orthogonally equivariant estimator of the population covariance matrix proposed by Tsai and Tsai (2024b) enjoys some optimal properties. Under some regularity conditions, they showed that their novel estimators of eigenvalues are consistent with the eigenvalues of the population covariance matrix.… ▽ More For the high-dimensional covariance estimation problem, when $\lim_{n\to \infty}p/n=c \in (0,1)$ the orthogonally equivariant estimator of the population covariance matrix proposed by Tsai and Tsai (2024b) enjoys some optimal properties. Under some regularity conditions, they showed that their novel estimators of eigenvalues are consistent with the eigenvalues of the population covariance matrix. In this note, first, we show that their novel estimator is a consistent estimator of the population covariance matrix under a high-dimensional asymptotic setup. Moreover, we may show that the novel estimator is the MLE of the population covariance matrix when $c \in (0, 1)$. The novel estimator is incorporated to establish the optimal decomposite $T_{T}^{2}-$test for a high-dimensional statistical hypothesis testing problem and to make the statistical inference for the high-dimensional principal component analysis-related problems without the sparsity assumption. Some remarks when $p >n $, especially for the high-dimensional low-sample size categorical data models $p >> n$, are made in the final section. △ Less

Submitted 3 November, 2024; v1 submitted 18 May, 2024; originally announced May 2024.

Comments: arXiv admin note: text overlap with arXiv:2405.06877

On the orthogonally equivariant estimators of a covariance matrix

Authors: Ming-Tien Tsai, Chia-Hsuan Tsai

Abstract: In this note, when the dimension $p$ is large we look into the insight of the Mar$\check{c}$enko-Pastur equation to get an explicit equality relationship, and use the obtained equality to establish a new kind of orthogonally equivariant estimator of the population covariance matrix. Under some regularity conditions, the proposed novel estimators of the population eigenvalues are shown to be consis… ▽ More In this note, when the dimension $p$ is large we look into the insight of the Mar$\check{c}$enko-Pastur equation to get an explicit equality relationship, and use the obtained equality to establish a new kind of orthogonally equivariant estimator of the population covariance matrix. Under some regularity conditions, the proposed novel estimators of the population eigenvalues are shown to be consistent for the eigenvalues of population covariance matrix. It is also shown that the proposed estimator is the best orthogonally equivariant estimator of population covariance matrix under the normalized Stein loss function. △ Less

Submitted 2 November, 2024; v1 submitted 10 May, 2024; originally announced May 2024.

The decomposite $T^{2}$-test when the dimension is large

Authors: Chia-Hsuan Tsai, Ming-Tien Tsai

Abstract: In this paper, we discuss tests for mean vector of high-dimensional data when the dimension $p$ is a function of sample size $n$. One of the tests, called the decomposite $T^{2}$-test, in the high-dimensional testing problem is constructed based on the estimation work of Ledoit and Wolf (2018), which is an optimal orthogonally equivariant estimator of the inverse of population covariance matrix un… ▽ More In this paper, we discuss tests for mean vector of high-dimensional data when the dimension $p$ is a function of sample size $n$. One of the tests, called the decomposite $T^{2}$-test, in the high-dimensional testing problem is constructed based on the estimation work of Ledoit and Wolf (2018), which is an optimal orthogonally equivariant estimator of the inverse of population covariance matrix under Stein loss function. The asymptotic distribution function of the test statistic is investigated under a sequence of local alternatives. The asymptotic relative efficiency is used to see whether a test is optimal and to perform the power comparisons of tests. An application of the decomposite $T^{2}$-test is in testing significance for the effect of monthly unlimited transport policy on public transportation, in which the data are taken from Taipei Metro System. △ Less

Submitted 3 March, 2024; originally announced March 2024.

Exact Homomorphic Encryption

Authors: Zheng-Yao Su, Ming-Chung Tsai

Abstract: Inspired by the concept of fault tolerance quantum computation, this article proposes a framework dubbed Exact Homomorphic Encryption, EHE, enabling exact computations on encrypted data without the need for pre-decryption. The introduction of quantum gates is a critical step for constructing the message encryption and the computation encryption within the framework. Of significance is that both en… ▽ More Inspired by the concept of fault tolerance quantum computation, this article proposes a framework dubbed Exact Homomorphic Encryption, EHE, enabling exact computations on encrypted data without the need for pre-decryption. The introduction of quantum gates is a critical step for constructing the message encryption and the computation encryption within the framework. Of significance is that both encryptions are respectively accomplished in a multivariate polynomial set generated by quantum gates. Two fundamental traits of quantum gates, the invertibility and the noncommutativity, establish the success of EHE. The encrypted computation is exact because its encryption transformation is conducted with invertible gates. In the same vein, decryptions for both an encrypted message and encrypted computation are exact. The second trait of noncommutativity among applied quantum gates brings forth the security for the two encryptions. Toward the message encryption, a plaintext is encoded into a ciphertext via a polynomial set generated by a product of noncommuting gates randomly chosen. In the computation encryption, a desired operation is encoded into an encrypted polynomial set generated by another product of noncommuting gates. The encrypted computation is then the evaluation of the encrypted polynomial set on the ciphertext and is referred to as the cryptovaluation. EHE is not only attainable on quantum computers, but also straightforwardly realizable on traditional computing environments. Surpassing the standard security 2^128 of quantum resilience, both the encryptions further reach a security greater than the suggested threshold 2^1024 and are characterized as hyper quantum-resilient. Thanks to the two essential traits of quantum gates, this framework can be regarded as the initial tangible manifestation of the concept noncommutative cryptography. △ Less

Submitted 8 May, 2024; v1 submitted 17 January, 2024; originally announced January 2024.

Superior visible photoelectric response with Au/Cu2NiSnS4 core-shell nanocrystals

Authors: Anima Ghosh, Shyam Narayan Singh Yadav, Ming-Hsiu Tsai, Abhishek Dubey, Shangjr Gwo, Chih-Ting Lin, Ta- Jen Yen

Abstract: The incorporation of plasmonic metal nanostructures into semiconducting chalcogenides, in the form of core-shell structures, represents a promising approach to boosting the performance of photodetectors. In this study, we combined Au nanoparticles with newly developed copper-based chalcogenides Cu2NiSnS4 (Au/CNTS), to achieve an ultrahigh optoelectronic response in the visible regime. The high-qua… ▽ More The incorporation of plasmonic metal nanostructures into semiconducting chalcogenides, in the form of core-shell structures, represents a promising approach to boosting the performance of photodetectors. In this study, we combined Au nanoparticles with newly developed copper-based chalcogenides Cu2NiSnS4 (Au/CNTS), to achieve an ultrahigh optoelectronic response in the visible regime. The high-quality Au/CNTS core-shell structure was synthesized by developing a unique colloidal hot-injection method, which allowed excellent control over sizes, shapes, and elemental compositions. The fabricated Au/CNTS hybrid core-shell structure exhibited enhanced optical absorption, carrier extraction efficiency, and improved photo-sensing performance, owing to the plasmonic-induced resonance energy transfer effect of the Au core. This effect led to a significant increase in carrier density between the Au core and CNTS shell. These values outperformed a CNTS-based gate-free visible photodetector. △ Less

Submitted 29 August, 2023; v1 submitted 6 August, 2023; originally announced August 2023.

Angle-of-Arrival Estimation of Narrow Gaussian Beams for Mobile FSO Platforms

Authors: Ming-Cheng Tsai, Muhammad Salman Bashir, Mohamed-Slim Alouini

Abstract: Due to the narrow beamwidths of laser Gaussian beams, accurate tracking of laser beam's angle-of-arrival is an important problem in mobile free-space optical communications. In most optical receivers today, fine tracking of angle-of-arrival involves estimating the location of the focused beam spot projected onto a focal plane array. However, for very thin Gaussian beams, both the location as well… ▽ More Due to the narrow beamwidths of laser Gaussian beams, accurate tracking of laser beam's angle-of-arrival is an important problem in mobile free-space optical communications. In most optical receivers today, fine tracking of angle-of-arrival involves estimating the location of the focused beam spot projected onto a focal plane array. However, for very thin Gaussian beams, both the location as well as the energy of the spot varies considerably with the variation of angle-of-arrival. In this study, we have analyzed the relationship between the angle-of-arrival and the energy of laser spot on the focal plane. We then exploited this relationship to enhance the angle-of-arrival estimation performance of our proposed receiver that takes into account both the location as well as the energy of the laser spot while estimating the angle-of-arrival. The derived Cramer-Rao bounds indicate that the system performance can be enhanced significantly for narrow Gaussian beams when both the spot location and energy are exploited for angle-of-arrival estimation. △ Less

Submitted 29 July, 2023; originally announced July 2023.

Study on the Correlation between Objective Evaluations and Subjective Speech Quality and Intelligibility

Authors: Hsin-Tien Chiang, Kuo-Hsuan Hung, Szu-Wei Fu, Heng-Cheng Kuo, Ming-Hsueh Tsai, Yu Tsao

Abstract: Subjective tests are the gold standard for evaluating speech quality and intelligibility; however, they are time-consuming and expensive. Thus, objective measures that align with human perceptions are crucial. This study evaluates the correlation between commonly used objective measures and subjective speech quality and intelligibility using a Chinese speech dataset. Moreover, new objective measur… ▽ More Subjective tests are the gold standard for evaluating speech quality and intelligibility; however, they are time-consuming and expensive. Thus, objective measures that align with human perceptions are crucial. This study evaluates the correlation between commonly used objective measures and subjective speech quality and intelligibility using a Chinese speech dataset. Moreover, new objective measures are proposed that combine current objective measures using deep learning techniques to predict subjective quality and intelligibility. The proposed deep learning model reduces the amount of training data without significantly affecting prediction performance. We analyzed the deep learning model to understand how objective measures reflect subjective quality and intelligibility. We also explored the impact of including subjective speech quality ratings on speech intelligibility prediction. Our findings offer valuable insights into the relationship between objective measures and human perceptions. △ Less

Submitted 10 October, 2023; v1 submitted 10 July, 2023; originally announced July 2023.

Improving Conversational Passage Re-ranking with View Ensemble

Authors: Jia-Huei Ju, Sheng-Chieh Lin, Ming-Feng Tsai, Chuan-Ju Wang

Abstract: This paper presents ConvRerank, a conversational passage re-ranker that employs a newly developed pseudo-labeling approach. Our proposed view-ensemble method enhances the quality of pseudo-labeled data, thus improving the fine-tuning of ConvRerank. Our experimental evaluation on benchmark datasets shows that combining ConvRerank with a conversational dense retriever in a cascaded manner achieves a… ▽ More This paper presents ConvRerank, a conversational passage re-ranker that employs a newly developed pseudo-labeling approach. Our proposed view-ensemble method enhances the quality of pseudo-labeled data, thus improving the fine-tuning of ConvRerank. Our experimental evaluation on benchmark datasets shows that combining ConvRerank with a conversational dense retriever in a cascaded manner achieves a good balance between effectiveness and efficiency. Compared to baseline methods, our cascaded pipeline demonstrates lower latency and higher top-ranking effectiveness. Furthermore, the in-depth analysis confirms the potential of our approach to improving the effectiveness of conversational search. △ Less

Submitted 26 April, 2023; originally announced April 2023.

Comments: SIGIR 2023