arXiv subjects

Geon Choi

Publications and source records attributed to Geon Choi.

15 recordsLinked to original sources

arXiv2026

Exploiting Structural Properties for Efficient Constraint-Aware HNSW Hyperparameter Tuning

Vector databases (VectorDBs) are a core component of modern retrieval systems, including Retrieval-Augmented Generation (RAG), where efficient Approximate Nearest Neighbor Search (ANNS) is critical. Among ANNS algorithms, Hierarchical Navigable Small World (HNSW) graphs are widely adopted for their strong recalllatency trade-off. However, configuring HNSW remains challenging: its hyperparameters jointly affect search quality, latency, build time, and index size in nonlinear ways, while production deployments impose strict resource and tuning-time constraints.We study HNSW hyperparameter tuning from a systems perspective and show that its configuration space exhibits strong structural regularities. Specifically, we identify monotonic, dominant unimodal, and separable relationships among search-time and construction-time parameters, which induce feasibility boundaries under performance and resource constraints. Building on this insight, we propose CHAT, a constraint-aware tuning framework for HNSW. Unlike generic black-box optimizers, CHAT exploits HNSW-specific structure to perform deterministic, sample-efficient search and prune resource-infeasible configurations before full index construction. Across multiple datasets and HNSW-based vector search engines, CHAT identifies configurations that maximize recall or throughput while satisfying constraints on accuracy, latency, build time, index size, and tuning budget. Compared to strong baselines, CHAT achieves up to 45% higher throughput or 11% higher recall, and converges up to 44x faster. These results show that principled, structure-aware tuning enables efficient and robust HNSW deployment beyond generic black-box optimization.

cs.DB↗

arXiv2026

CheXpercept: A Benchmark for Evaluating Expert-Level Lesion Perception in Chest X-rays

The evaluation of vision-language models (VLMs) for chest X-ray (CXR) analysis has largely been limited to disease-presence classification without visual grounding. Such evaluations fail to verify the expert-level lesion perception necessary to ensure the clinical reliability of VLMs. To address these limitations, we introduce CheXpercept, a sequential, multi-level perception benchmark that mirrors a radiologist's cognitive workflow across coarse-level detection, fine-level contour evaluation and revision, and semantic-level attribute extraction. To ensure high clinical fidelity at scale, we construct the dataset using a semi-automated generation pipeline paired with a review by six medical experts. CheXpercept contains 10,400 QA items derived from 2,100 CXRs, covering seven clinically critical pulmonary and cardiac lesions. To demonstrate the current landscape of VLM perception, we benchmark 14 general and medical VLMs on CheXpercept. The models achieve adequate performance only at the coarse level, with accuracy degrading precipitously on deeper visual tasks. Notably, medical VLMs show almost no perceptual advantage over their general-domain counterparts, highlighting a systemic flaw in current domain adaptation. The code and dataset will be publicly available.

cs.CV↗

arXiv2026

Learning While Transmitting: Pilotless Polar Coded Modulation for Short Packet Transmission

Short packets make channel learning expensive. In pilot-aided transmission (PAT), a non-negligible fraction of the packet is consumed by pilots, creating a direct pre-log loss and tightening the reliability margin needed for ultra-reliable low-latency communication. We propose a pilot-free polar-coded framework that replaces explicit pilots with \emph{coded pilots}. The message is carried by two polar-coded segments: a quadrature phase shift keying (QPSK) segment that is decodable without channel state information (CSI), and a higher-order quadrature amplitude modulation (QAM) segment that provides high spectral efficiency. The receiver employs \emph{hybrid decoding}: it first jointly infers CSI during successive-cancellation-based decoding of the QPSK segment by exploiting QPSK phase-rotation invariance together with polar frozen-bit constraints; the decoded QPSK symbols then act as \emph{implicit pilots} for coherent detection and decoding of the QAM segment. The split also makes rate adaptation practical by confining the symmetry/frozen-bit requirements for phase resolution to the QPSK segment, enabling puncturing and shortening without breaking the pilot-free mechanism. For multi-block fading, we optimize the split and code parameters via density evolution with Gaussian approximation (DEGA); for higher-order modulation, we use bit-interleaved coded modulation capacity approximation to obtain equivalent channel parameters. Incorporating channel-estimation error variance into the DEGA-based analysis, simulations over practical multi-block block-fading channels show gains up to $1.5$~dB over PAT in the short-blocklength regime.

eess.SP↗

arXiv2026

Hierarchical Subcode Ensemble Decoding of Polar Codes

Subcode-ensemble decoders improve iterative decoding by running multiple decoders in parallel over carefully chosen subcodes, increasing the likelihood that at least one decoder avoids the dominant trapping structures. Achieving strong diversity gains, however, requires constructing many subcodes that satisfy a linear covering property-yet existing approaches lack a systematic way to scale the ensemble size while preserving this property. This paper introduces hierarchical subcode ensemble decoding (HSCED), a new ensemble decoding framework that expands the number of constituent decoders while still guaranteeing linear covering. The key idea is to recursively generate subcode parity constraints in a hierarchical structure so that coverage is maintained at every level, enabling large ensembles with controlled complexity. To demonstrate its effectiveness, we apply HSCED to belief propagation (BP) decoding of polar codes, where dense parity-check matrices induce severe stopping-set effects that limit conventional BP. Simulations confirm that HSCED delivers significant block-error-rate improvements over standard BP and conventional subcode-ensemble decoding under the same decoding-latency constraint.

cs.IT↗

arXiv2026

Multipoint Code-Weight Sphere Decoding: Parallel Near-ML Decoding for Short-Blocklength Codes

Ultra-reliable low-latency communications (URLLC) operate with short packets, where finite-blocklength effects make near-maximum-likelihood (near-ML) decoding desirable but often too costly. This paper proposes a two-stage near-ML decoding framework that applies to any linear block code. In the first stage, we run a low-complexity decoder to produce a candidate codeword and a cyclic redundancy check. When this stage succeeds, we terminate immediately. When it fails, we invoke a second-stage decoder, termed multipoint code-weight sphere decoding (MP-WSD). The central idea behind {MP-WSD} is to concentrate the ML search where it matters. We pre-compute a set of low-weight codewords and use them to generate structured local perturbations of the current estimate. Starting from the first-stage output, MP-WSD iteratively explores a small Euclidean sphere of candidate codewords formed by adding selected low-weight codewords, tightening the search region as better candidates are found. This design keeps the average complexity low: at high signal-to-noise ratio, the first stage succeeds with high probability and the second stage is rarely activated; when it is activated, the search remains localized. Simulation results show that the proposed decoder attains near-ML performance for short-blocklength, low-rate codes while maintaining low decoding latency.

cs.IT↗

arXiv2025

Instruction-Guided Lesion Segmentation for Chest X-rays with Automatically Generated Large-Scale Dataset

The applicability of current lesion segmentation models for chest X-rays (CXRs) has been limited both by a small number of target labels and the reliance on complex, expert-level text inputs, creating a barrier to practical use. To address these limitations, we introduce instruction-guided lesion segmentation (ILS), a medical-domain adaptation of referring image segmentation (RIS) designed to segment diverse lesion types based on simple, user-friendly instructions. Under this task, we construct MIMIC-ILS, the first large-scale instruction-answer dataset for CXR lesion segmentation, using our fully automated multimodal pipeline that generates annotations from CXR images and their corresponding reports. MIMIC-ILS contains 1.1M instruction-answer pairs derived from 192K images and 91K unique segmentation masks, covering seven major lesion types. To empirically demonstrate its utility, we present ROSALIA, a LISA model fine-tuned on the MIMIC-ILS dataset. ROSALIA can segment diverse lesions and provide textual explanations in response to user instructions. The model achieves high accuracy in our newly proposed task, highlighting the effectiveness of our pipeline and the value of MIMIC-ILS as a foundational resource for pixel-level CXR lesion grounding. The dataset and model are available at https://github.com/checkoneee/ROSALIA.

cs.CV↗

arXiv2025

Code-Weight Sphere Decoding

Ultra-reliable low-latency communications (URLLC) demand high-performance error-correcting codes and decoders in the finite blocklength regime. This letter introduces a novel two-stage near-maximum likelihood (near-ML) decoding framework applicable to any linear block code. Our approach first employs a low-complexity initial decoder. If this initial stage fails a cyclic redundancy check, it triggers a second stage: the proposed code-weight sphere decoding (WSD). WSD iteratively refines the codeword estimate by exploring a localized sphere of candidates constructed from pre-computed low-weight codewords. This strategy adaptively minimizes computational overhead at high signal-to-noise ratios while achieving near-ML performance, especially for low-rate codes. Extensive simulations demonstrate that our two-stage decoder provides an excellent trade-off between decoding reliability and complexity, establishing it as a promising solution for next-generation URLLC systems.

eess.SP↗

arXiv2025

Lunguage: A Benchmark for Structured and Sequential Chest X-ray Interpretation

Radiology reports convey detailed clinical observations and capture diagnostic reasoning that evolves over time. However, existing evaluation methods are limited to single-report settings and rely on coarse metrics that fail to capture fine-grained clinical semantics and temporal dependencies. We introduce LUNGUAGE, a benchmark dataset for structured radiology report generation that supports both single-report evaluation and longitudinal patient-level assessment across multiple studies. It contains 1,473 annotated chest X-ray reports, each reviewed by experts, and 186 of them contain longitudinal annotations to capture disease progression and inter-study intervals, also reviewed by experts. Using this benchmark, we develop a two-stage structuring framework that transforms generated reports into fine-grained, schema-aligned structured reports, enabling longitudinal interpretation. We also propose LUNGUAGESCORE, an interpretable metric that compares structured outputs at the entity, relation, and attribute level while modeling temporal consistency across patient timelines. These contributions establish the first benchmark dataset, structuring framework, and evaluation metric for sequential radiology reporting, with empirical results demonstrating that LUNGUAGESCORE effectively supports structured report evaluation. The code is available at: https://github.com/SuperSupermoon/Lunguage

cs.CL↗

arXiv2025

CXReasonBench: A Benchmark for Evaluating Structured Diagnostic Reasoning in Chest X-rays

Recent progress in Large Vision-Language Models (LVLMs) has enabled promising applications in medical tasks, such as report generation and visual question answering. However, existing benchmarks focus mainly on the final diagnostic answer, offering limited insight into whether models engage in clinically meaningful reasoning. To address this, we present CheXStruct and CXReasonBench, a structured pipeline and benchmark built on the publicly available MIMIC-CXR-JPG dataset. CheXStruct automatically derives a sequence of intermediate reasoning steps directly from chest X-rays, such as segmenting anatomical regions, deriving anatomical landmarks and diagnostic measurements, computing diagnostic indices, and applying clinical thresholds. CXReasonBench leverages this pipeline to evaluate whether models can perform clinically valid reasoning steps and to what extent they can learn from structured guidance, enabling fine-grained and transparent assessment of diagnostic reasoning. The benchmark comprises 18,988 QA pairs across 12 diagnostic tasks and 1,200 cases, each paired with up to 4 visual inputs, and supports multi-path, multi-stage evaluation including visual grounding via anatomical region selection and diagnostic measurements. Even the strongest of 12 evaluated LVLMs struggle with structured reasoning and generalization, often failing to link abstract knowledge with anatomically grounded visual interpretation. The code is available at https://github.com/ttumyche/CXReasonBench

cs.CV↗

arXiv2025

Rate-Matching Deep Polar Codes via Polar Coded Extension

Deep polar codes are pre-transformed polar codes that employ a multi-layered polar kernel transformation strategy to enhance code performance in short blocklength regimes. However, like conventional polar codes, their block length is constrained to powers of two, as the final transformation layer uses a conventional polar kernel matrix. This paper introduces a novel rate-matching technique for deep polar codes using code extension, particularly effective when the desired code length slightly exceeds a power of two. The key idea is to exploit the layered structure of deep polar codes by concatenating polar codewords generated at each transformation layer. Based on this structure, we also develop an efficient decoding algorithm leveraging soft-output successive cancellation list decoding and provide comprehensive error probability analysis supporting our code design algorithms. Additionally, we propose a computationally efficient greedy algorithm for multi-layer configurations. Extensive simulations confirm that our approach delivers substantial coding gains over conventional rate-matching methods, especially in medium to high code-rate regimes.

cs.IT↗

arXiv2024

Sparsely Pre-transformed Polar Codes for Low-Latency SCL Decoding

Deep polar codes, employing multi-layered polar kernel pre-transforms in series, are recently introduced variants of pre-transformed polar codes. These codes have demonstrated the ability to reduce the number of minimum weight codewords, thereby closely achieving finite-block length capacity with successive cancellation list (SCL) decoders in certain scenarios. However, when the list size of the SCL decoder is small, which is crucial for low-latency communication applications, the reduction in the number of minimum weight codewords does not necessarily improve decoding performance. To address this limitation, we propose an alternative pre-transform technique to enhance the suitability of polar codes for SCL decoders with practical list sizes. Leveraging the fact that the SCL decoding error event set can be decomposed into two exclusive error event sets, our approach applies two different types of pre-transformations, each targeting the reduction of one of the two error event sets. Extensive simulation results under various block lengths and code rates have demonstrated that our codes consistently outperform all existing state-of-the-art pre-transformed polar codes, including CRC-aided polar codes and polarization-adjusted convolutional codes, when decoded using SCL decoders with small list sizes.

cs.IT↗

arXiv2024

Nonlinear Self-Interference Cancellation With Learnable Orthonormal Polynomials for Full-Duplex Wireless Systems

Nonlinear self-interference cancellation (SIC) is essential for full-duplex communication systems, which can offer twice the spectral efficiency of traditional half-duplex systems. The challenge of nonlinear SIC is similar to the classic problem of system identification in adaptive filter theory, whose crux lies in identifying the optimal nonlinear basis functions for a nonlinear system. This becomes especially difficult when the system input has a non-stationary distribution. In this paper, we propose a novel algorithm for nonlinear digital SIC that adaptively constructs orthonormal polynomial basis functions according to the non-stationary moments of the transmit signal. By combining these basis functions with the least mean squares (LMS) algorithm, we introduce a new SIC technique, called as the adaptive orthonormal polynomial LMS (AOP-LMS) algorithm. To reduce computational complexity for practical systems, we augment our approach with a precomputed look-up table, which maps a given modulation and coding scheme to its corresponding basis functions. Numerical simulation indicates that our proposed method surpasses existing state-of-the-art SIC algorithms in terms of convergence speed and mean squared error when the transmit signal is non-stationary, such as with adaptive modulation and coding. Experimental evaluation with a wireless testbed confirms that our proposed approach outperforms existing digital SIC algorithms.

eess.SP↗

arXiv2023

Deep Polar Codes

In this paper, we introduce a novel class of pre-transformed polar codes, termed as deep polar codes. We first present a deep polar encoder that harnesses a series of multi-layered polar transformations with varying sizes. Our approach to encoding enables a low-complexity implementation while significantly enhancing the weight distribution of the code. Moreover, our encoding method offers flexibility in rate-profiling, embracing a wide range of code rates and blocklengths. Next, we put forth a low-complexity decoding algorithm called successive cancellation list with backpropagation parity checks (SCL-BPC). This decoding algorithm leverages the parity check equations in the reverse process of the multi-layered pre-transformed encoding for SCL decoding. Additionally, we present a low-latency decoding algorithm that employs parallel-SCL decoding by treating partially pre-transformed bit patterns as additional frozen bits. Through simulations, we demonstrate that deep polar codes outperform existing pre-transformed polar codes in terms of block error rates across various code rates under short block lengths, while maintaining low encoding and decoding complexity. Furthermore, we show that concatenating deep polar codes with cyclic-redundancy-check codes can achieve the meta-converse bound of the finite block length capacity within 0.4 dB in some instances.

cs.IT↗

arXiv2023

Information-Theoretical Approach to Integrated Pulse-Doppler Radar and Communication Systems

Integrated sensing and communication improves the design of systems by combining sensing and communication functions for increased efficiency, accuracy, and cost savings. The optimal integration requires understanding the trade-off between sensing and communication, but this can be difficult due to the lack of unified performance metrics. In this paper, an information-theoretical approach is used to design the system with a unified metric. A sensing rate is introduced to measure the amount of information obtained by a pulse-Doppler radar system. An approximation and lower bound of the sensing rate is obtained in closed forms. Using both the derived sensing information and communication rates, the optimal bandwidth allocation strategy is found for maximizing the weighted sum of the spectral efficiency for sensing and communication. The simulation results confirm the validity of the approximation and the effectiveness of the proposed bandwidth allocation.

cs.IT↗

arXiv2022

Split-KalmanNet: A Robust Model-Based Deep Learning Approach for SLAM

Simultaneous localization and mapping (SLAM) is a method that constructs a map of an unknown environment and localizes the position of a moving agent on the map simultaneously. Extended Kalman filter (EKF) has been widely adopted as a low complexity solution for online SLAM, which relies on a motion and measurement model of the moving agent. In practice, however, acquiring precise information about these models is very challenging, and the model mismatch effect causes severe performance loss in SLAM. In this paper, inspired by the recently proposed KalmanNet, we present a robust EKF algorithm using the power of deep learning for online SLAM, referred to as Split-KalmanNet. The key idea of Split-KalmanNet is to compute the Kalman gain using the Jacobian matrix of a measurement function and two recurrent neural networks (RNNs). The two RNNs independently learn the covariance matrices for a prior state estimate and the innovation from data. The proposed split structure in the computation of the Kalman gain allows to compensate for state and measurement model mismatch effects independently. Numerical simulation results verify that Split-KalmanNet outperforms the traditional EKF and the state-of-the-art KalmanNet algorithm in various model mismatch scenarios.

eess.SP↗