Selected work

Representative Papers

Learning joint probabilistic weather forecasts from station observations alone

Oct 07, 2026

This study addresses the challenge of generating joint probabilistic weather forecasts with inter-variable dependencies using only station observations to assess compound meteorological risks. To this end, we propose CLARA, a lightweight, CPU-friendly architecture comprising approximately 28,000 parameters. By leveraging a calibrated advection-routing attention mechanism, CLARA directly learns joint Gaussian predictive distributions for five surface variables from station data without requiring numerical weather predictions or reanalysis products. Furthermore, we develop a consistent covariance scale estimator and demonstrate that neglecting inter-variable correlations significantly degrades negative log-likelihood performance. Experiments across ten global regions reveal that CLARA reduces energy scores by 4.9%–65% relative to baselines, outperforming the persistence baseline in all 60 comparisons and a comparable-scale model in 57 instances.

0 citationsRead paper

POLAR: Ontology-Guided Risk Prevention for Tool-Calling LLM Agents

Oct 06, 2026

This study addresses the absence of preemptive risk prevention mechanisms in LLM tool agents by proposing POLAR, a novel defense framework. POLAR introduces a dual-layer ontology-based reversibility grading score and a candidate inverse sequence derivation mechanism. By leveraging structured ontological modeling to assess operation reversibility, the framework prunes high-risk tool calls prior to execution and supports integration with small-model agents. Experimental evaluations on τ²-bench demonstrate that POLAR effectively improves task rewards for specific scenarios while delineating the trade-off boundary between utility and safety. Ultimately, this work provides auditable, preemptive safety guardrails for LLM agents, mitigating risks before irreversible actions occur.

0 citationsRead paper

Shared Stopping Decisions Change Answers in HQQ Cache Quantization

Oct 05, 2026

This study addresses the cross-request coupling introduced by global error-based stopping mechanisms in HQQ-quantized KV caches during batched inference, where irrelevant requests interfere with target outputs. By analyzing the error propagation pathways inherent in shared stopping logic, this work proposes fixed-iteration and local stopping strategies to decouple intra-batch dependencies. Crucially, it reveals that quantized stopping decisions, rather than quantization itself, constitute the root cause of answer drift. This research establishes a novel perspective emphasizing the necessity of auditing stopping logic beyond merely optimizing quantization group configurations. Experimental results confirm that the proposed approach effectively eliminates companion-induced output shifts, while also demonstrating that natural rebatching may still alter model responses, thereby offering essential guidance for the safe deployment of HQQ in production systems.

0 citationsRead paper

Early warning of road icing from antecedent meteorological conditions without consistent critical slowing down

Sep 27, 2026

This study investigates the efficacy of critical slowing down (CSD) signals for early warning of road icing under rapid external forcing. Utilizing one-minute high-frequency road measurement data from South Korea, we conduct multi-station validation through variance and lag-1 autocorrelation analyses alongside ROC evaluation. We provide the first empirical evidence that conventional CSD indicators fail to reliably detect icing precursors in environmental transitions dominated by external forcing. Accordingly, we propose a novel early warning framework integrating recent meteorological forcing with cumulative conditions. Results demonstrate that incorporating extended meteorological fields improves the AUC by 0.032 and reduces false alarm rates by over 62% within 3- to 6-hour prediction windows. This work effectively redefines the focal points of early warning systems for such rapidly forced scenarios.

0 citationsRead paper

Physics-residual machine learning predicts oxygen-evolution catalyst activity beyond the training range from sparse polarization measurements

Sep 20, 2026

通过物理残差机器学习方法，从稀疏极化测量中预测超出训练范围的氧演化催化剂活性，提高筛选效率。

0 citationsRead paper

Recent publications

Latest Papers

Learning joint probabilistic weather forecasts from station observations alone

Oct 07, 2026

This study addresses the challenge of generating joint probabilistic weather forecasts with inter-variable dependencies using only station observations to assess compound meteorological risks. To this end, we propose CLARA, a lightweight, CPU-friendly architecture comprising approximately 28,000 parameters. By leveraging a calibrated advection-routing attention mechanism, CLARA directly learns joint Gaussian predictive distributions for five surface variables from station data without requiring numerical weather predictions or reanalysis products. Furthermore, we develop a consistent covariance scale estimator and demonstrate that neglecting inter-variable correlations significantly degrades negative log-likelihood performance. Experiments across ten global regions reveal that CLARA reduces energy scores by 4.9%–65% relative to baselines, outperforming the persistence baseline in all 60 comparisons and a comparable-scale model in 57 instances.

0 citationsRead paper

POLAR: Ontology-Guided Risk Prevention for Tool-Calling LLM Agents

Oct 06, 2026

This study addresses the absence of preemptive risk prevention mechanisms in LLM tool agents by proposing POLAR, a novel defense framework. POLAR introduces a dual-layer ontology-based reversibility grading score and a candidate inverse sequence derivation mechanism. By leveraging structured ontological modeling to assess operation reversibility, the framework prunes high-risk tool calls prior to execution and supports integration with small-model agents. Experimental evaluations on τ²-bench demonstrate that POLAR effectively improves task rewards for specific scenarios while delineating the trade-off boundary between utility and safety. Ultimately, this work provides auditable, preemptive safety guardrails for LLM agents, mitigating risks before irreversible actions occur.

0 citationsRead paper

Shared Stopping Decisions Change Answers in HQQ Cache Quantization

Oct 05, 2026

This study addresses the cross-request coupling introduced by global error-based stopping mechanisms in HQQ-quantized KV caches during batched inference, where irrelevant requests interfere with target outputs. By analyzing the error propagation pathways inherent in shared stopping logic, this work proposes fixed-iteration and local stopping strategies to decouple intra-batch dependencies. Crucially, it reveals that quantized stopping decisions, rather than quantization itself, constitute the root cause of answer drift. This research establishes a novel perspective emphasizing the necessity of auditing stopping logic beyond merely optimizing quantization group configurations. Experimental results confirm that the proposed approach effectively eliminates companion-induced output shifts, while also demonstrating that natural rebatching may still alter model responses, thereby offering essential guidance for the safe deployment of HQQ in production systems.

0 citationsRead paper

Early warning of road icing from antecedent meteorological conditions without consistent critical slowing down

Sep 27, 2026

This study investigates the efficacy of critical slowing down (CSD) signals for early warning of road icing under rapid external forcing. Utilizing one-minute high-frequency road measurement data from South Korea, we conduct multi-station validation through variance and lag-1 autocorrelation analyses alongside ROC evaluation. We provide the first empirical evidence that conventional CSD indicators fail to reliably detect icing precursors in environmental transitions dominated by external forcing. Accordingly, we propose a novel early warning framework integrating recent meteorological forcing with cumulative conditions. Results demonstrate that incorporating extended meteorological fields improves the AUC by 0.032 and reduces false alarm rates by over 62% within 3- to 6-hour prediction windows. This work effectively redefines the focal points of early warning systems for such rapidly forced scenarios.

0 citationsRead paper

Physics-residual machine learning predicts oxygen-evolution catalyst activity beyond the training range from sparse polarization measurements

Sep 20, 2026

通过物理残差机器学习方法，从稀疏极化测量中预测超出训练范围的氧演化催化剂活性，提高筛选效率。

0 citationsRead paper