Searcharxiv⌕ Search

arXiv subjects

Minwoo Kim

Publications and source records attributed to Minwoo Kim.

At least 19 recordsLinked to original sources
arXiv2026

FinRED: An Expert-Guided Benchmark Generation and Evaluation Framework for Financial LLM Red-Teaming

Existing safety benchmarks target general adversarial scenarios but miss finance-specific risks. Financial LLMs face regulatory compliance violations, fraud facilitation, and systemic trust erosion that require targeted evaluation. We introduce FinRED, an expert-guided red-teaming framework for financial LLM safety evaluation developed with financial experts. FinRED uses a novel two-level taxonomy mapping global standards (e.g., FATF and EU DORA) to threats ranging from regulatory evasion to complex fraud, integrated with a scalable pipeline that converts real financial documents into context-rich red-teaming Behavioral Prompts (seeds) through an expert-defined schema. Rigorous expert validation confirms seed plausibility and realism for meaningful LLM safety evaluation. We also provide an expert-validated, finance-specific rubric that goes beyond disclaimer checks, aligns more closely with human experts than static one-size-fits-all rubrics, and reduces critical false negatives from 28 to 12. Aligned with internationally adopted risk-management and information-security standards (e.g., ISO/IEC 27001), FinRED is deployed in South Korea's Financial Security Institute (FSI) regulatory sandbox for generative AI security evaluation in real financial services. To mitigate dual-use risks, the dataset, generation pipeline, prompt template, and evaluation framework are gated for qualified researchers at https://github.com/selectstar-ai/FinRED-paper and https://huggingface.co/datasets/datumo/FinRED.

cs.CR↗
arXiv2026

Nuclear quantum effects enhance diffusion in supercooled ammonia through accelerated pyramidal inversion

In liquid ammonia, an imbalance between available hydrogen-bond donor and acceptor sites limits water-like tetrahedral connectivity, leaving sparse and short-lived associations within a densely packed liquid. We ask whether quantum-sensitive local rearrangements can nevertheless contribute appreciably to diffusion in this weakly connected liquid. We compare classical and thermostatted ring polymer molecular dynamics from 170 to 250 K using an r$^2$SCAN-trained machine learning force field. Ammonia can change its molecular geometry as the nitrogen atom passes through the plane of the three hydrogen atoms, a motion known as pyramidal inversion. Nuclear quantum effects increase the inversion rate and reduce its apparent activation energy. Inversion events are accompanied by transient reductions in hydrogen-bond coordination and local density within the first solvation shell. Although nuclear quantum effects lower density and viscosity throughout the studied range, diffusion remains only weakly affected in the warmer liquid. The quantum enhancement of diffusion emerges below approximately 210 K and reaches 18% at 170 K, tracking the enhancement of cage escape. These observations support an inversion-assisted contribution to diffusion, in which a spatially localized, quantum-sensitive internal motion couples to cage relaxation without a persistent tetrahedral network.

physics.chem-ph↗
arXiv2026

Guard Models Are Overconfident Where Base Models Are Uncertain

Guard models are used as safety classifiers, with confidence scores driving downstream moderation decisions. We evaluate five guard models for prompt classification and find that although several are nearly calibrated on clean inputs, adversarial attacks degrade their calibration by an order of magnitude, turning false negatives into high-confidence errors indistinguishable from correct detections. Comparing each guard with its corresponding base LM, we find that uncertainty signals often remain available, with the base model typically expressing uncertainty on the same inputs where the guard fails. Layer-wise analyses localize this guard-base divergence to later layers, where guard models exhibit sharper safe/unsafe separation and lower-rank representations, while adversarial harmful inputs lie closer to the clean-safe region. These findings highlight a mismatch between guard confidence and base model uncertainty under attack.

cs.LG↗
arXiv2026

LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models

Vision-Language-Action (VLA) models achieve strong performance in robotic manipulation by leveraging pre-trained vision-language backbones. However, in downstream robotic settings, they are typically fine-tuned with limited data, leading to overfitting to specific instruction formulations and leaving robustness to paraphrased instructions underexplored. To study this gap, we introduce LIBERO-Para, a controlled benchmark that independently varies action expressions and object references for fine-grained analysis of linguistic generalization. Across seven VLA configurations (0.6B-7.5B), we observe consistent performance degradation of 22-52 pp under paraphrasing. This degradation is primarily driven by object-level lexical variation: even simple synonym substitutions cause large drops, indicating reliance on surface-level matching rather than semantic grounding. Moreover, 80-96% of failures arise from planning-level trajectory divergence rather than execution errors, showing that paraphrasing disrupts task identification. Binary success rate treats all paraphrases equally, obscuring whether models perform consistently across difficulty levels or rely on easier cases. To address this, we propose PRIDE, a metric that quantifies paraphrase difficulty using semantic and syntactic factors. Our benchmark and corresponding code are available at: https://github.com/cau-hai-lab/LIBERO-Para

cs.LG↗
arXiv2026

FRESHLATENT: Channel-Aware Latent Adaptation for Resource-Constrained Embodied VLM Perception

Mission-critical UAVs increasingly rely on split vision-language model (VLM) perception under tight onboard-resource and wireless-communication constraints. However, corruption of transmitted intermediate features creates a deployment mismatch for clean-trained split interfaces, while stronger channel-aware codecs can impose substantial onboard cost. We present FreshLatent, a lightweight channel-aware latent adapter that trains a power-normalized encoder-decoder through wireless corruption while keeping the surrounding VLM frozen. We formulate deployment around a mission-conditioned perception requirement and embedded interface cost, linking channel quality and communication budget to the operating conditions under which perception remains usable. At 0 dB and the tightest communication budget, FreshLatent improves gIoU and cIoU over clean split compression by 20.79 and 20.87 points, respectively. At the most adverse evaluated SNR (0 dB), across all three communication budgets, FreshLatent recovers 63.5-69.1% of the gIoU improvement achieved by a much heavier, range-trained feature-JSCC codec. On an NVIDIA Jetson AGX Xavier in 10-W mode, FreshLatent uses 37-40x fewer encoder parameters, 7.7-9.9x lower edge-interface latency, and 8.8-10.0x lower edge-interface energy than the heavier codec. Together, these results show that lightweight channel-aware adaptation can recover a substantial fraction of the robustness of a much larger communication interface while broadening quality-valid operation under constrained wireless conditions.

eess.SP↗
arXiv2026

Prediction-Aided V2X Safety Message Recovery via Uncertainty-Aware LDPC Decoding

Periodic basic safety messages (BSMs) exhibit temporal correlation as vehicle motion evolves continuously over time. This temporal structure provides predictive information about the current BSM and motivates prediction-aided recovery after a decoding failure. For low-density parity-check (LDPC)-coded transmission, a straightforward approach maps a predicted vehicle state to a candidate bit sequence and uses it as decoder prior information. However, a point prediction does not capture the uncertainty of the prediction. The resulting hard point prior assigns fixed confidence to each predicted bit and may introduce strong erroneous evidence when the prediction is incorrect. This paper proposes an uncertainty-aware prediction-aided LDPC recovery framework for vehicle-to-everything (V2X) safety messages. Instead of using a deterministic point prediction, the proposed method represents the current vehicle state by a predictive distribution. The distribution is propagated through BSM field quantization and serialization to obtain bit-level probabilities, which are converted into prior log-likelihood ratios (LLRs) whose magnitudes reflect prediction reliability. After an initial cyclic redundancy check (CRC) failure, these probabilistic priors are combined with the original channel observations and used in a second LDPC decoding pass. The proposed method preserves the standardized message representation and channel-coding procedure. Simulation results show that the proposed probabilistic priors outperform the hard point prior across different motion predictors. At $E_b/N_0=0.75$~dB, the proposed method recovers 87.27% of the messages that fail the initial decoding attempt.

eess.SP↗
arXiv2026

Token-Oriented Semantic Communication with Pretrained Vision Transformers

Token communications realize the semantic communication principle at the granularity of transformer tokens, providing a promising direction for client--server collaborative inference in resource-constrained edge systems. However, directly transmitting token embeddings presents two practical challenges: substantial communication cost and limited interoperability across model-specific token embedding spaces. To address these challenges, we propose a \emph{token-oriented} semantic communication framework. In this framework, token-level task relevance determines which compressed image latents are transmitted, enabling token-granular transmission without directly transmitting token embeddings. The framework is modular, coordinating three pretrained components---a lightweight client-side vision transformer (ViT), a learned image compression (LIC) model, and a large server-side ViT---without end-to-end training. The key enabler is the one-to-one spatial alignment between ViT patch tokens and the LIC latent vectors, which allows token-level task relevance to directly determine which latent vectors are transmitted. Building on this alignment, token-aligned LIC selectively transmits task-relevant latents, layer-selective attention rollout estimates token relevance from a selected range of attention layers in a single forward pass, and surrogate token substitution adapts the frozen server model by optimizing a single learnable token. Experiments on ImageNet show that the proposed framework achieves a more favorable rate--accuracy trade-off than recent semantic communication schemes, hand-crafted codecs, and task-agnostic LIC models.

eess.SP↗
arXiv2026

FRAME: Factored Retrieval via Attribute Readouts for Object-Centric Scene Memory

Language-guided robots need persistent scene memories to follow instructions, revisit objects, and resolve references to objects encountered over time. While much of language-guided scene-memory retrieval has emphasized spatial or relational references, many everyday object references specify objects by multiple persistent attributes, such as category, material, size, or surface appearance. We formalize this problem as attribute-compositional retrieval, where a fixed object-centric scene memory is queried with natural language to retrieve the object satisfying the requested attributes. To investigate this capability directly, we introduce a controlled evaluation protocol with fixed scene memories and attribute-defined targets, separating retrieval from perception and annotation ambiguities. We then propose FRAME, which turns language into query-relevant attribute weights, uses learned readouts to estimate per-attribute evidence from object embeddings, and ranks objects by aggregating this evidence according to the query. Across held-out scenes and object assets, FRAME outperforms representative scene-memory retrieval baselines while reducing post-decomposition object scoring to lightweight matrix-vector computation. These results position attribute-compositional retrieval as a complementary scene-memory capability for language-guided robots, showing that persistent object attributes can be exposed as composable evidence for accurate and efficient multi-attribute retrieval.

cs.CV↗
arXiv2026

LUCID: An Agentic AI Framework on Digital-Twin in the Loop for QoS-Guaranteeing Robotic Control

Cloud robotics relies on the timely uplink of high-volume sensing streams, yet dynamic environments continually shift the feasible combinations of trajectories, active-robot count, and per-robot QoS. Because existing approaches formulate trajectory planning (TP) and radio resource management (RRM) as a single fixed optimization problem, they cannot reconfigure these coupled decisions as conditions evolve, resulting in transient QoS violations. However, evolving operator intents change which quantities-such as the active-robot count and per-robot QoS-are fixed, optimized, or relaxed. Furthermore, the computational cost of evaluating trajectory-dependent wireless conflicts has made it difficult to build large-scale Digital-Twin-in-the-Loop (DITL) testbeds responsive enough for such dynamic orchestration. We present LUCID, an LLM-agent--orchestrated, uplink-aware cloud-robotics pipeline that moves TP--RRM from solving a fixed formulation to dynamically orchestrating optimization problem schemas within a DITL environment. Driven by the operator's high-level intent, LUCID treats the TP--RRM formulation as a bounded template whose variables, objectives, and constraints are dynamically configured, while SimBridge enables repeated ray-tracing evaluation by converting large-scale robotics scenes into wireless-ready DTs. By integrating collision-free path planning with a spectral-radius RRM validator, LUCID identifies wireless bottlenecks and restructures the problem schema on the fly to efficiently find the verified feasible state. Experiments confirm that LUCID robustly adapts to changing intents, active-robot counts, and scenes, while a multimodal surrogate model, FastConfigNet, reduces planning latency.

eess.SY↗
arXiv2026

Real-time feedback control of ELM frequency using divertor gas puffing and its effects on tungsten-induced radiation and plasma performance in KSTAR

The edge-localized mode (ELM) frequency ($f_{\mathrm{ELM}}$) was successfully controlled in real time on KSTAR using a proportional-integral (PI) feedback controller, employing a $\mathrm{D}_2$ divertor gas puff as the actuator under tungsten lower-divertor conditions. The controller accurately tracked a two-step target---a 30 Hz increase in $f_{\mathrm{ELM}}$ for 4 s, followed by a 30 Hz decrease for 3 s---yielding mean and median absolute percentage errors of approximately 13% and 12%, respectively. Compared to a reference discharge, the actively controlled shot did not exhibit a significant drop in volume-integrated core radiation, confirming that excessive gas use merely degrades overall plasma performance. However, when contrasted with the exponential increase in core radiation observed in the absence of divertor gas puffing, these results underscore the critical need for real-time optimization. Specifically, divertor gas commands must be actively managed to maintain an $f_{\mathrm{ELM}}$ sufficient for flushing tungsten from the core while maximizing global plasma performance.

physics.plasm-ph↗
arXiv2026

SSDi8: Accurate and Efficient 8-bit Quantization for State Space Duality

Recent advances in sequence modeling have highlighted Mamba as a state space architecture offering efficient long-range dependency modeling and providing a viable alternative to Transformers. Building upon this, Mamba-2 introduces the Structured State Space Duality (SSD), which integrates recurrent and attention modes to achieve efficiency and scalability. However, this architectural expansion substantially increases memory and latency overhead, underscoring the need for efficient compression strategies tailored to SSD. In this work, we present SSDi8, the first post-training quantization framework specifically designed for SSD to maintain a persistent INT8 path. SSDi8 introduces a reformulation that decouples element-wise multiplications from matrix multiplications, enabling reuse of quantized activations across modules. Moreover, SSDi8 adaptively quantizes channel-varying activations at cost-effective points, further reducing latency. On the accuracy side, SSDi8 explicitly leverages the intrinsic dimensional decomposition of SSD, exploiting distinct outlier distributions across axes, and incorporates an error correction term based on per-channel error statistics. Comprehensive experiments demonstrate that SSDi8 achieves accuracy comparable to FP16 while delivering up to 1.4x speedup in W4A8 and W8A8 settings. We further validate its robustness in resource-constrained environments by deploying it on the Orin NX device.

cs.AI↗
arXiv2026

KorNAT: LLM Alignment Benchmark for Korean Social Values and Common Knowledge

For Large Language Models (LLMs) to be effectively deployed in a specific country, they must possess an understanding of the nation's culture and basic knowledge. To this end, we introduce National Alignment, which measures an alignment between an LLM and a targeted country from two aspects: social value alignment and common knowledge alignment. Social value alignment evaluates how well the model understands nation-specific social values, while common knowledge alignment examines how well the model captures basic knowledge related to the nation. We constructed KorNAT, the first benchmark that measures national alignment with South Korea. For the social value dataset, we obtained ground truth labels from a large-scale survey involving 6,174 unique Korean participants. For the common knowledge dataset, we constructed samples based on Korean textbooks and GED reference materials. KorNAT contains 4K and 6K multiple-choice questions for social value and common knowledge, respectively. Our dataset creation process is meticulously designed and based on statistical sampling theory and was refined through multiple rounds of human review. The experiment results of seven LLMs reveal that only a few models met our reference score, indicating a potential for further enhancement. KorNAT has received government approval after passing an assessment conducted by a government-affiliated organization dedicated to evaluating dataset quality. Samples and detailed evaluation protocols of our dataset can be found in https://huggingface.co/datasets/jiyounglee0523/KorNAT and https://github.com/jiyounglee-0523/KorNAT.

cs.CL↗
arXiv2026

FloodReasonBench: Benchmarking VLM Reasoning Segmentation for Embodied Flood Response at the Edge

Reasoning segmentation enables vision-language models (VLMs) to translate mission-relevant language requests into pixel-level visual grounding, offering a natural perception interface for embodied agents. However, existing benchmarks largely focus on generic visual scenes and overlook the domain and resource constraints encountered in flood-response platforms. We present FloodReasonBench, a benchmark for VLM reasoning segmentation for embodied flood response at the edge. At its core, FloodReasonBench introduces FloodResponseSeg, a flood-specific reasoning-segmentation dataset constructed from real-world scenes and response-relevant targets. Beyond task accuracy, the benchmark characterizes reasoning-segmentation pipelines under lightweight visual encoding, hierarchical split inference, and compressed intermediate representations. We observe strong partition-dependent accuracy variation in the generic pre-adaptation setting, while the flood-adapted target-workload design space exhibits a substantially more compact accuracy range across partitions. Evaluation on an NVIDIA Jetson AGX Xavier further exposes the tradeoffs among reasoning-segmentation accuracy, edge-side latency, energy, and communication footprint, enabling quality-constrained selection of edge operating points. Together, these results provide a task- and system-level characterization of reasoning segmentation for resource-constrained embodied flood response at the edge.

cs.DC↗
arXiv2026

ImpactHO: Importance-Aware KV Cache Transfer for Multi-User Edge LLM Handover

Edge LLMs must preserve inference continuity when a user hands over between edge nodes, requiring key-value (KV) cache transfer to the target node. However, simultaneous handovers saturate the backhaul, preventing full cache delivery within the mobility-imposed transfer window. Rather than allocating bandwidth as if all cache entries were equally valuable, we order each user's KV cache by importance and transmit only its most informative fraction, turning token-level sparsity into communication savings. We cast the transfer as a multi-user backhaul allocation problem that maximizes average accuracy across users. Each user's partial-cache accuracy serves as its utility: a sigmoid that fits measurements on the RULER benchmark with $R^2>0.99$ across models and context lengths. Because importance ordering front-loads the high-value entries, the concave region of the accuracy curve spans nearly the entire cache. Our proposed allocator keeps served users within this region, making each per-slot allocation problem convex. The optimum is derived via a closed-form weighted water-filling solution that generalizes information-theoretic water-filling and enables online scheduling. The proposed allocator attains over 93.7% average accuracy in a 500ms transfer window, within 0.5pp of the full-cache ceiling, and reaches 98.2-99.5% of a clairvoyant upper bound.

cs.NI↗
arXiv2026

Prune Once: Retraining-Free Task-Agnostic Pruning for Vision-Language Models

Vision-language models (VLMs) have achieved remarkable generalization across diverse multimodal tasks through large-scale pre-training, yet their rapidly increasing computational and memory requirements pose significant challenges for deployment in constrained environments. Existing pruning strategies often depend on task-specific criteria or LLM-oriented importance measures, making them unsuitable for task-agnostic pruning, where no task-specific samples are available at pruning time and the pruned model remains broadly applicable. We introduce a retraining-free VLM pruning framework called PORTA that derives a task- and modality-agnostic importance formulation based on activation variation, estimated from generic calibration data, which reliably captures feature-level representation utility across modalities. PORTA further incorporates an adaptive sparsity allocation mechanism that assigns layer-wise pruning ratios based on output feature variability, avoiding the limitations of uniform sparsity and reducing performance degradation at high compression levels. Extensive experiments across VLM architectures, such as CLIP, BLIP, and Qwen2-VL, demonstrate that PORTA achieves competitive downstream performance under high sparsity without requiring any retraining, supporting efficient VLM compression. Code is available at https://github.com/cau-hai-lab/PORTA.git.

cs.CV↗
arXiv2026

SceneBaker: Radio-Ready Scene Generation for Sionna Ray-Tracing

Wireless digital-twin (DT) research needs ray-tracing (RT) scenes that can be generated, versioned, and checked reproducibly. Current visual-authoring workflows can produce plausible city models, but they are poorly matched to repeated radio-simulation studies because geometry, terrain contact, and material semantics often require manual repair after export. This paper presents SceneBaker, a programmatic scene-generation pipeline that turns map and terrain data into Sionna RT-ready Mitsuba scenes without a GUI authoring step. Across four campus scenes, SceneBaker generates Sionna RT-ready scenes over the same geographic bounds as a Blender-generated baseline. The generated scenes avoid representative building-generation and terrain-contact faults while preserving comparable coverage-field and link-level channel-response behavior. The implementation and generated comparison assets are available at https://github.com/hslyu/sionna-scene-baker.

eess.SP↗
arXiv2026

Temporal Channel Estimation for Generalized CSI Feedback

Efficient Channel State Information (CSI) feedback is indispensable for frequency division duplex (FDD) massive multiple-input multiple-output (MIMO) systems. Existing compressed sensing (CS) algorithms exploit delay-domain sparsity but suffer from prohibitive iterative latency and discrete grid mismatch. Conversely, deep learning (DL) approaches achieve rapid inference but lack spatial scalability and domain adaptability, failing to generalize to unseen propagation environments, and demand computationally heavy encoders and decoder. In this paper, we propose TAP, a Tap-Assisted Parametric CSI Compression. TAP is a one-shot neural framework that unifies the speed of DL with the mathematical interpretability of CS. TAP replaces iterative pursuit with a lightweight 1D neural network that extracts dominant continuous propagation delays from temporal channel sequences via a differentiable sub-grid interpolation operator. TAP achieves true architecture independence, enabling zero-shot generalization across diverse array geometries and unseen propagation environments. Furthermore, TAP yields a completely decoder-free payload, allowing the BS to reconstruct the channel via a simple inverse fast Fourier transform (IFFT). Extensive evaluations across five 3GPP environments demonstrate that TAP achieves a 3.13 to 12.22 dB channel frequency response normalized mean square error (CFR-NMSE) improvement over CsiNet while shrinking the model footprint by 660 times to under 1 MB. Operating with sub-millisecond latencies, TAP accelerates inference by 2700 times over classical iterative OMP, providing a scalable and deployment-ready solution for next-generation networks.

eess.SP↗
arXiv2026

Joint estimation of high-dimensional spiked covariance matrices via a partially shared subspace

Statistical analysis of high-dimensional data is often hampered by limited sample sizes, yet auxiliary datasets from related sources are often readily available. When two such datasets share part of their covariance structure, but not all of it, exploiting the shared part can substantially improve estimation. We propose a spiked covariance model that explicitly captures this partial sharing: two datasets share a subspace of unknown rank and arbitrary position in the spectrum, while each retains its own distinct spiked directions. The model treats the two datasets symmetrically and strictly generalizes existing models for shared covariance structure. We develop a complete estimation procedure that includes joint estimation of the shared subspace and its rank, a closed-form pooling weight for combining the two datasets, and asymptotic guarantees derived from random matrix theory in the proportional-growth regime. The framework also resolves a gap in contrastive dimension reduction by providing a principled estimator for high-dimensional settings. We illustrate the methodology on portfolio construction during the early COVID-19 pandemic and on contrastive analysis of brain tumor gene expression.

stat.ME↗