Selected work

Representative Papers

An Exploration Graph with Continuous Refinement for Efficient Multimedia Retrieval

Jul 29, 2026

This work addresses the challenges faced by existing graph-based methods in large-scale multimedia approximate nearest neighbor search (ANNS), namely slow construction speed, high memory overhead, and difficulty in balancing exploratory retrieval performance. To overcome these limitations, we propose the continuously refined Exploration Graph (crEG), which introduces—for the first time—an undirected even-degree connected graph structure specifically designed for exploratory retrieval. This structure ensures real-time connectivity while enabling optional edge optimization. By integrating rapid graph construction, a continuous refinement strategy, and an edge optimization algorithm, crEG significantly improves construction efficiency and memory utilization without compromising retrieval accuracy, outperforming state-of-the-art ANNS methods on exploratory retrieval tasks.

0 citationsRead paper

Dynamic Exploration Graph: A Novel Approach for Efficient Nearest Neighbor Search in Evolving Multimedia Datasets

Jul 29, 2026

This work addresses the challenge of balancing efficiency and structural stability in graph-based approximate nearest neighbor search under dynamic multimedia scenarios involving frequent data insertions and deletions. The authors propose the Dynamic Exploration Graph (DEG), which introduces a novel vertex deletion strategy that preserves graph connectivity and a distribution-agnostic expansion mechanism to maintain high search accuracy and efficiency in evolving environments. Experimental results demonstrate that DEG significantly outperforms existing dynamic graph-based methods on dynamic datasets, achieving superior performance in both index construction time and query efficiency. Moreover, DEG attains state-of-the-art results even on static datasets, confirming its robustness and general applicability.

0 citationsRead paper

Fast and Efficient Approximate Nearest Neighbor Search for High-Dimensional LLM Embeddings

Jul 23, 2026

This work addresses the challenge of efficient approximate nearest neighbor search and maximum inner product search (MIPS) in high-dimensional embeddings from large language models. The authors propose a novel approach that transforms asymmetric MIPS into Euclidean nearest neighbor search via dimension augmentation, combined with Equi-Voronoi Polytopes (EVP) quantization and a Fast Linear Assignment Sorting (FLAS) one-dimensional pre-sorting mechanism. This integration substantially accelerates k-nearest neighbor graph (kNNG) construction and query processing while enhancing memory access locality and cache efficiency. Evaluated in the SISAP 2026 challenge, the method achieves low-latency, high-recall MIPS performance, significantly outperforming existing baselines.

0 citationsRead paper

AI-Augmented Human Resource Management? Insights from German companies

Jul 15, 2026

This study investigates the application of generative AI and predictive analytics in human resource management within German firms, examining the tension between efficiency gains and strategic talent development. Drawing on a mixed-methods approach—including 410 survey responses, in-depth interviews, and focus groups—the research reveals that AI plays a dual role in HR: enhancing predictive capabilities while reinforcing an efficiency-oriented mindset. Findings indicate that AI significantly augments HR’s data analytics capacity, yet its deployment remains predominantly focused on process optimization. The study further underscores the critical influence of data governance, algorithmic transparency, and organizational transformation on the successful implementation of AI in HR contexts, offering both theoretical insights and practical guidance for developing AI-driven human resource strategies.

0 citationsRead paper

ConceptTracer: Interactive Analysis of Concept Saliency and Selectivity in Neural Representations

Apr 08, 2026

This work addresses the lack of systematic tools for uncovering the relationship between representations learned by neural networks—particularly tabular foundation models—and human-interpretable concepts. To bridge this gap, we introduce ConceptTracer, the first interactive analysis framework that integrates information-theoretic measures of concept saliency and selectivity, specifically designed for tabular foundation models such as TabPFN. By quantifying both the strength and specificity of individual neuron responses to predefined concepts and coupling these metrics with an intuitive visual interface, ConceptTracer enables users to efficiently identify neurons with clear semantic interpretations. Experiments on TabPFN successfully reveal neurons whose activations align closely with human-understandable concepts, demonstrating the effectiveness and practical utility of our approach in elucidating how neural networks encode conceptual knowledge.

0 citationsRead paper

Recent publications

Latest Papers

An Exploration Graph with Continuous Refinement for Efficient Multimedia Retrieval

Jul 29, 2026

This work addresses the challenges faced by existing graph-based methods in large-scale multimedia approximate nearest neighbor search (ANNS), namely slow construction speed, high memory overhead, and difficulty in balancing exploratory retrieval performance. To overcome these limitations, we propose the continuously refined Exploration Graph (crEG), which introduces—for the first time—an undirected even-degree connected graph structure specifically designed for exploratory retrieval. This structure ensures real-time connectivity while enabling optional edge optimization. By integrating rapid graph construction, a continuous refinement strategy, and an edge optimization algorithm, crEG significantly improves construction efficiency and memory utilization without compromising retrieval accuracy, outperforming state-of-the-art ANNS methods on exploratory retrieval tasks.

0 citationsRead paper

Dynamic Exploration Graph: A Novel Approach for Efficient Nearest Neighbor Search in Evolving Multimedia Datasets

Jul 29, 2026

This work addresses the challenge of balancing efficiency and structural stability in graph-based approximate nearest neighbor search under dynamic multimedia scenarios involving frequent data insertions and deletions. The authors propose the Dynamic Exploration Graph (DEG), which introduces a novel vertex deletion strategy that preserves graph connectivity and a distribution-agnostic expansion mechanism to maintain high search accuracy and efficiency in evolving environments. Experimental results demonstrate that DEG significantly outperforms existing dynamic graph-based methods on dynamic datasets, achieving superior performance in both index construction time and query efficiency. Moreover, DEG attains state-of-the-art results even on static datasets, confirming its robustness and general applicability.

0 citationsRead paper

Fast and Efficient Approximate Nearest Neighbor Search for High-Dimensional LLM Embeddings

Jul 23, 2026

This work addresses the challenge of efficient approximate nearest neighbor search and maximum inner product search (MIPS) in high-dimensional embeddings from large language models. The authors propose a novel approach that transforms asymmetric MIPS into Euclidean nearest neighbor search via dimension augmentation, combined with Equi-Voronoi Polytopes (EVP) quantization and a Fast Linear Assignment Sorting (FLAS) one-dimensional pre-sorting mechanism. This integration substantially accelerates k-nearest neighbor graph (kNNG) construction and query processing while enhancing memory access locality and cache efficiency. Evaluated in the SISAP 2026 challenge, the method achieves low-latency, high-recall MIPS performance, significantly outperforming existing baselines.

0 citationsRead paper

AI-Augmented Human Resource Management? Insights from German companies

Jul 15, 2026

This study investigates the application of generative AI and predictive analytics in human resource management within German firms, examining the tension between efficiency gains and strategic talent development. Drawing on a mixed-methods approach—including 410 survey responses, in-depth interviews, and focus groups—the research reveals that AI plays a dual role in HR: enhancing predictive capabilities while reinforcing an efficiency-oriented mindset. Findings indicate that AI significantly augments HR’s data analytics capacity, yet its deployment remains predominantly focused on process optimization. The study further underscores the critical influence of data governance, algorithmic transparency, and organizational transformation on the successful implementation of AI in HR contexts, offering both theoretical insights and practical guidance for developing AI-driven human resource strategies.

0 citationsRead paper

ConceptTracer: Interactive Analysis of Concept Saliency and Selectivity in Neural Representations

Apr 08, 2026

This work addresses the lack of systematic tools for uncovering the relationship between representations learned by neural networks—particularly tabular foundation models—and human-interpretable concepts. To bridge this gap, we introduce ConceptTracer, the first interactive analysis framework that integrates information-theoretic measures of concept saliency and selectivity, specifically designed for tabular foundation models such as TabPFN. By quantifying both the strength and specificity of individual neuron responses to predefined concepts and coupling these metrics with an intuitive visual interface, ConceptTracer enables users to efficiently identify neurons with clear semantic interpretations. Experiments on TabPFN successfully reveal neurons whose activations align closely with human-understandable concepts, demonstrating the effectiveness and practical utility of our approach in elucidating how neural networks encode conceptual knowledge.

0 citationsRead paper