arXiv subjects

Robert Chang

Publications and source records attributed to Robert Chang.

9 recordsLinked to original sources

arXiv2025

Benchmarking Next-Generation Reasoning-Focused Large Language Models in Ophthalmology: A Head-to-Head Evaluation on 5,888 Items

Recent advances in reasoning-focused large language models (LLMs) mark a shift from general LLMs toward models designed for complex decision-making, a crucial aspect in medicine. However, their performance in specialized domains like ophthalmology remains underexplored. This study comprehensively evaluated and compared the accuracy and reasoning capabilities of four newly developed reasoning-focused LLMs, namely DeepSeek-R1, OpenAI o1, o3-mini, and Gemini 2.0 Flash-Thinking. Each model was assessed using 5,888 multiple-choice ophthalmology exam questions from the MedMCQA dataset in zero-shot setting. Quantitative evaluation included accuracy, Macro-F1, and five text-generation metrics (ROUGE-L, METEOR, BERTScore, BARTScore, and AlignScore), computed against ground-truth reasonings. Average inference time was recorded for a subset of 100 randomly selected questions. Additionally, two board-certified ophthalmologists qualitatively assessed clarity, completeness, and reasoning structure of responses to differential diagnosis questions.O1 (0.902) and DeepSeek-R1 (0.888) achieved the highest accuracy, with o1 also leading in Macro-F1 (0.900). The performance of models across the text-generation metrics varied: O3-mini excelled in ROUGE-L (0.151), o1 in METEOR (0.232), DeepSeek-R1 and o3-mini tied for BERTScore (0.673), DeepSeek-R1 (-4.105) and Gemini 2.0 Flash-Thinking (-4.127) performed best in BARTScore, while o3-mini (0.181) and o1 (0.176) led AlignScore. Inference time across the models varied, with DeepSeek-R1 being slowest (40.4 seconds) and Gemini 2.0 Flash-Thinking fastest (6.7 seconds). Qualitative evaluation revealed that DeepSeek-R1 and Gemini 2.0 Flash-Thinking tended to provide detailed and comprehensive intermediate reasoning, whereas o1 and o3-mini displayed concise and summarized justifications.

cs.CL↗

arXiv2025

Scaling IP Lookup to Large Databases using the CRAM Lens

Wide-area scaling trends require new approaches to Internet Protocol (IP) lookup, enabled by modern networking chips such as Intel Tofino, AMD Pensando, and Nvidia BlueField, which provide substantial ternary content-addressable memory (TCAM) and static random-access memory (SRAM). However, designing and evaluating scalable algorithms for these chips is challenging due to hardware-level constraints. To address this, we introduce the CRAM (CAM+RAM) lens, a framework that combines a formal model for evaluating algorithms on modern network processors with a set of optimization idioms. We demonstrate the effectiveness of CRAM by designing and evaluating three new IP lookup schemes: RESAIL, BSIC, and MashUp. RESAIL enables Tofino-2 to scale to 2.25 million IPv4 prefixes$\unicode{x2014}$likely sufficient for the next decade$\unicode{x2014}$while a pure TCAM approach supports only 250k prefixes, just 27% of the current global IPv4 routing table. Similarly, BSIC scales to 390k IPv6 prefixes on Tofino-2, supporting 3.2 times as many prefixes as a pure TCAM implementation. In contrast, existing state-of-the-art algorithms, SAIL for IPv4 and Hi-BST for IPv6, scale to considerably smaller sizes on Tofino-2.

cs.NI↗

arXiv2023

Extraction of Text from Optic Nerve Optical Coherence Tomography Reports

Purpose: The purpose of this study was to develop and evaluate rule-based algorithms to enhance the extraction of text data, including retinal nerve fiber layer (RNFL) values and other ganglion cell count (GCC) data, from Zeiss Cirrus optical coherence tomography (OCT) scan reports. Methods: DICOM files that contained encapsulated PDF reports with RNFL or Ganglion Cell in their document titles were identified from a clinical imaging repository at a single academic ophthalmic center. PDF reports were then converted into image files and processed using the PaddleOCR Python package for optical character recognition. Rule-based algorithms were designed and iteratively optimized for improved performance in extracting RNFL and GCC data. Evaluation of the algorithms was conducted through manual review of a set of RNFL and GCC reports. Results: The developed algorithms demonstrated high precision in extracting data from both RNFL and GCC scans. Precision was slightly better for the right eye in RNFL extraction (OD: 0.9803 vs. OS: 0.9046), and for the left eye in GCC extraction (OD: 0.9567 vs. OS: 0.9677). Some values presented more challenges in extraction, particularly clock hours 5 and 6 for RNFL thickness, and signal strength for GCC. Conclusions: A customized optical character recognition algorithm can identify numeric results from optical coherence scan reports with high precision. Automated processing of PDF reports can greatly reduce the time to extract OCT results on a large scale.

eess.IV↗

arXiv2022

Szeg\H{o} kernel asymptotics and concentration of Husimi Distributions of eigenfunctions

We work on the boundary $\partial M_\tau$ of a Grauert tube of a closed, real analytic Riemannian manifold $M$. The Toeplitz operator $\Pi_\tau D_{\sqrt{\rho}} \Pi_\tau$ associated to the Reeb vector field is a positive, self-adjoint, elliptic operator on $H^2(\partial M_\tau)$. We compute $\lambda \to \infty$ asymptotics under parabolic rescaling in a neighborhood of the geodesic (Reeb) flow $G^{t}_{\tau} = \exp t\Xi_{\sqrt{\rho}}$ for the spectral projection kernel $\Pi_{\chi, \lambda}$ associated to $\Pi_\tau D_{\sqrt{\rho}} \Pi_\tau$. We also compute scaling asymptotics for tempered sums of Husimi distributions (analytic continuations) on $\partial M_\tau$ of Laplace eigenfunctions on $M$. Both asymptotic formulae can be expressed in terms of the metaplectic representation of the linearization of the geodesic flow $G^{t}_\tau$ on Bargmann--Fock space. As a corollary, we obtain sharp $L^p \to L^{q}$ norm estimates for $\Pi_{\chi, \lambda}$ and sharp $L^p$ estimates for Husimi distributions.

math.SP↗

arXiv2021

SoK: A Study of the Security on Voice Processing Systems

As the use of Voice Processing Systems (VPS) continues to become more prevalent in our daily lives through the increased reliance on applications such as commercial voice recognition devices as well as major text-to-speech software, the attacks on these systems are increasingly complex, varied, and constantly evolving. With the use cases for VPS rapidly growing into new spaces and purposes, the potential consequences regarding privacy are increasingly more dangerous. In addition, the growing number and increased practicality of over-the-air attacks have made system failures much more probable. In this paper, we will identify and classify an arrangement of unique attacks on voice processing systems. Over the years research has been moving from specialized, untargeted attacks that result in the malfunction of systems and the denial of services to more general, targeted attacks that can force an outcome controlled by an adversary. The current and most frequently used machine learning systems and deep neural networks, which are at the core of modern voice processing systems, were built with a focus on performance and scalability rather than security. Therefore, it is critical for us to reassess the developing voice processing landscape and to identify the state of current attacks and defenses so that we may suggest future developments and theoretical improvements.

cs.CR↗

arXiv2021

Scaling asymptotics for Szeg\H{o} kernels on Grauert tubes

Let $M_\tau$ be the Grauert tube of radius $\tau$ of a closed, real analytic manifold $M$. Associated to the Grauert tube boundary is the orthogonal projection $\Pi_\tau \colon L^2(\partial M_\tau) \to H^2(\partial M_\tau)$, called the Szeg\H{o} projector. Let $D_{\sqrt{\rho}}$ denote the Hamilton vector field of the Grauert tube function $\sqrt{\rho}$ acting as a differential operator. We prove scaling asymptotics for the spectral localization kernel of the Toeplitz operator $\Pi_\tau D_{\sqrt{\rho}} \Pi_\tau$. We also prove scaling asymptotics for the tempered spectral projections kernel $P_{\chi, \lambda}(z,w) = \sum_{\lambda_j \le \lambda} e^{-2\tau\lambda_j} \phi_{\lambda_j}^\mathbb{C}(z) \overline{\phi_{\lambda_j}^\mathbb{C}(w)}$, where $\phi_{\lambda_j}^\mathbb{C}$ are analytic extensions to the Grauert tube of Laplace eigenfunctions on $M$.

math.SP↗

arXiv2018

Log-scale equidistribution of nodal sets in Grauert tubes

Let $M_{τ_0}$ be the Grauert tube (of some fixed radius $τ_0$) of a compact, negatively curved, real analytic Riemannian manifold $M$ without boundary. Let $ϕ_λ$ be a Laplacian eigenfunction on $M$ of eigenvalues $-λ^2$ and let $ϕ_λ^\mathbb{C}$ be its holomorphic extension to $M_{τ_0}$. In this article, we prove that on $M_{τ_0} \setminus M$, there exists a dimensional constant $α> 0$ and a full density subsequence $ \{λ_{j_k}\}_{k=1}^{\infty}$ of the spectrum for which the masses of the complexified eigenfunctions $ϕ_{λ_{j_k}}^\mathbb{C}$ are asymptotically equidistributed at length scale $(\log λ_{j_k})^{-α}$. Moreover, the complex zeros of $ϕ_{λ_{j_k}}^\mathbb{C}$ also become equidistributed on this logarithmic length scale.

math.AP↗

arXiv2018

Log-scale equidistribution of zeros of quantum ergodic eigensections

Under suitable hypotheses, a symplectic map can be quantized as a sequence of unitary operators acting on the $N$th powers of a positive line bundle over a Kähler manifold. We show that if the symplectic map has polynomial decay of correlations, then there exists a density one subsequence of eigensections whose masses and zeros become equidistributed in balls of logarithmically shrinking radii of lengths $\lvert \log N \rvert^{-γ}$ for some constant $γ> 0$ independent of $N$.

math.SP↗

arXiv2015

Quantum ergodicity of Wigner induced spherical harmonics

We show that a Wigner induced random orthonormal basis of spherical harmonics is almost surely quantum ergodic. Here, a random basis is identified with an element of the product probability space of unitary groups, each endowed with the measure induced by the generalized Wigner ensemble. This yields a semi-classical realization of the probabilistic quantum unique ergodicity result for Wigner eigenvectors of Bourgade-Yau. At the same time, this generalizes a similar result due to Zelditch, who uses Haar measure on the unitary groups in defining the notion of a random basis.

math.PR↗