Selected work

Representative Papers

On Ray Reordering Techniques for Faster GPU Ray Tracing

May 04, 2020ACM Symposium on Interactive 3D Graphics and Games

Secondary rays in GPU-accelerated ray tracing exhibit poor spatial locality and low SIMT execution efficiency, limiting overall performance. Method: This paper proposes a hardware-agnostic ray reordering framework centered on an Endpoint-Estimated Key—designed specifically for secondary rays—to enhance spatial coherence with minimal computational overhead. The approach integrates wavefront path tracing scheduling, cooperative utilization of RTX hardware traversal kernels, and GPU memory access locality modeling, without relying on shader-specific implementations. Contribution/Results: Evaluated on modern GPUs, the method achieves 1.3–2.0× end-to-end ray tracing speedup. Reordering significantly accelerates software-intensive stages (e.g., intersection computation and shading preparation); although hardware traversal incurs inherent overhead, the net performance gain remains unambiguous. This work establishes a novel, portable paradigm for general-purpose GPU ray reordering—enabling efficient, implementation-independent acceleration across diverse ray tracing pipelines.

18 citations3 influentialRead paper

Intuitive Surgical SurgToolLoc Challenge Results: 2022-2023

May 11, 2023

To address the challenge of real-time, robust surgical instrument localization in minimally invasive robotic-assisted surgery (RAS) video streams, this work introduces SurgToolLoc—the first large-scale, multi-view, multi-scenario benchmark dataset with pixel-level mask annotations. We further propose a novel evaluation protocol emphasizing both cross-center generalizability and real-time inference (≥30 FPS). Methodologically, we integrate instance segmentation and keypoint detection with temporal modeling (ConvLSTM/Transformer), domain adaptation, and weakly supervised learning. Our best-performing model achieves 92.4% mAP@0.5 on the test set while maintaining an inference speed of 36 FPS—substantially outperforming conventional template matching and early CNN-based approaches. The solution has undergone rigorous preclinical validation across multiple surgical scenarios. By providing a reproducible, scalable, end-to-end framework for visual instrument localization in RAS, this work establishes a new standard for benchmarking and advancing vision-based surgical navigation systems.

17 citationsRead paper

Q3DE: A fault-tolerant quantum computer architecture for multi-bit burst errors by cosmic rays

Oct 01, 2022Micro

Multi-bit burst errors (MBBEs) induced by cosmic rays severely compromise the scalability of fault-tolerant quantum computing. Method: This paper proposes Q3DE, a low-overhead fault-tolerance enhancement architecture built within the surface code framework. Its core innovation is the first syndrome-based, transparent MBBE detection mechanism, integrated with dynamic logical encoding reconstruction and rollback-aware decoding—enabling real-time anomaly identification, decoding rollback, and recovery operation re-evaluation without hardware redundancy. Contribution/Results: By jointly optimizing dynamic code deformation and decoding, Q3DE reduces MBBE duration by 1000× and shrinks the affected qubit region by 50%, substantially alleviating stringent constraints on physical qubit density and chip footprint. This establishes a new paradigm for designing highly reliable, large-scale quantum processors.

15 citations3 influentialRead paper

Towards Understanding Retrieval Accuracy and Prompt Quality in RAG Systems

Nov 29, 2024arXiv.org

The impact of key design decisions—RAG activation, retrieval granularity, and knowledge integration strategy—on RAG system performance remains poorly understood. Method: We conduct systematic ablation studies across three code/qa benchmarks and two state-of-the-art LLMs, quantitatively evaluating how document type, recall rate, document selection strategy, and prompt engineering jointly affect answer correctness and confidence via multi-dimensional analysis, cross-model/dataset comparison, and joint prompt-retrieval analysis. Contribution/Results: We identify precise interaction patterns and operational boundaries among these factors and propose nine actionable, empirically grounded guidelines for diagnosing and optimizing RAG failures. Our findings significantly improve RAG system stability, debuggability, and reliability, offering rigorous empirical evidence and a principled methodology to support the engineering deployment of LLM-augmented systems.

7 citationsRead paper

Accurate Forgetting for Heterogeneous Federated Continual Learning

Feb 20, 2025International Conference on Learning Representations

To address statistical bias and noise interference arising from client data/task heterogeneity—or even adversarial behavior—in federated continual learning (FCL), this paper introduces the “Accurate Forgetting” (AF) paradigm: proactively identifying and discarding unreliable feature representations induced by skewed distributions and noise prior to knowledge reuse. Methodologically, we propose the first probability-based credibility assessment framework built upon normalized flows, enabling quantifiable, knowledge-granular filtering. Further, we integrate generative replay with selective knowledge inheritance to dynamically enhance global model robustness within the federated architecture. Evaluated on multiple heterogeneous FCL benchmarks, AF achieves an average accuracy improvement of 12.3%, significantly boosting generalization and noise resilience. Our approach provides a novel, interpretable, and computationally tractable pathway for bias mitigation in FCL.

5 citationsRead paper

Recent publications

Latest Papers

Distribution Steering via Sliced Optimal Transport Control

Aug 13, 2026

This work addresses the problem of designing feedback control laws to steer the state distribution of a dynamical system from a given initial distribution to a prescribed terminal distribution. Building upon sliced optimal transport, the authors propose a finite-horizon sliced feedback control framework that defines directional terminal conditions via one-dimensional projections and combines minimum-energy control with directional averaging to construct deterministic feedback laws, thereby circumventing the need for high-dimensional optimal transport maps. The approach preserves the affine structure for Gaussian distributions, enabling exact regulation of both mean and covariance, and incorporates distribution-dependent gains that guarantee linear decay of the sliced Wasserstein distance. Theoretical analysis shows that the proposed stochastic controller converges to the mean sliced flow as the sampling period tends to zero, and the framework naturally extends to general linear systems, with numerical experiments confirming its efficacy.

0 citationsRead paper

Impact of introducing "Informatics I" to the common university entrance examination in Japan: a longitudinal study on students' perceptions of their information-related knowledge and skills from 2006 to 2026

Aug 13, 2026

This study investigates the potential distorting effects of high-stakes standardized testing on the coverage of information education content and students’ metacognitive self-assessments. Leveraging longitudinal survey data from University of Tokyo freshmen spanning 2006–2026 and drawing on the backwash effect framework, it disentangles the distinct impacts of the 2013 and 2022 curriculum reforms and the 2025 inclusion of “Information I” in the national university entrance examination. By innovatively integrating long-term panel data with a backwash analytical lens, the research finds that the 2013 reform did not induce a cognitive discontinuity, whereas the impending 2025 exam change significantly increases students’ self-reported exposure to computer science content. However, this shift stems primarily from altered judgment criteria rather than genuine gains in competence, thereby revealing the constitutive role of assessment regimes in shaping learners’ self-perceptions.

0 citationsRead paper

Validation-Centric AI-Assisted GPU Porting of a 250,000+ Line Legacy Weather Simulation Code

Aug 13, 2026

This work addresses the tension between performance gains and scientific validity when porting large legacy scientific codes to GPUs by proposing a verification-centric, AI-assisted migration workflow. The approach integrates a large language model–driven CLI agent, OpenACC-based automated code transformation, and physics-informed kernel benchmark generation, ensuring consistency through both element-wise numerical comparison and application-level meteorological simulations. For the first time, scientific validation is deeply embedded into an AI-assisted porting pipeline, enabling automatic detection of floating-point semantic discrepancies and branch sensitivity, while highlighting the critical roles of conversational context management and runtime state reconstruction. Applied to the 250K-line Fortran weather model CReSS, the method successfully produced verified GPU implementations for 162 core kernels, achieving a 5.1× speedup in real typhoon simulations and uncovering five instances of numerical divergence, substantially reducing migration costs.

0 citationsRead paper

NARU: A Benchmark for NARrative Evolution and Cultural Nuance Understanding in Japanese Extreme Long Video

Aug 13, 2026

Existing video understanding benchmarks struggle to evaluate models’ ability to jointly comprehend narrative progression and culturally embedded meanings in high-context, non-English long-form videos. To address this gap, this work introduces the first large-scale benchmark for Japanese ultra-long videos, comprising 155 videos (146.8 hours) and 1,481 questions, uniquely integrating narrative tracking with cultural reasoning. The authors propose a hierarchical memory-driven annotation pipeline, a task-oriented question synthesis mechanism, and a two-stage native-speaker validation protocol coupled with iterative shortcut elimination to ensure cognitive depth and question quality. Evaluations reveal that current multimodal large language models exhibit significant limitations in long-range narrative integration and cultural inference.

0 citationsRead paper

Measuring Task-Agnostic Training Data Influence Across Language Model Pretraining

Aug 13, 2026

This work addresses the challenge of consistently measuring data influence during language model pretraining, a task hindered by existing methods’ reliance on specific downstream tasks or validation sets and their need for frequent evaluation of intermediate checkpoints. The authors propose a task-agnostic influence metric that quantifies how closely the gradient update from an individual training sample approximates the trajectory of final model parameters. This approach enables efficient estimation using intermediate checkpoints without requiring full retraining. For the first time, it facilitates data influence tracking throughout pretraining. Validated across 18 configurations of Pythia and PolyPythia models, the method reveals a systematic pattern: literary data exerts dominant influence early in training, while STEM-related data gains prominence later—a trend consistent across diverse model settings.

0 citationsRead paper