The record is part of the task: matched-record evaluation of text classifiers across maintenance, safety and recall reporting
研究通过比较同一案例在不同阶段的记录，评估文本分类器在维护、安全和召回报告中的表现差异。
0 citationsRead paper
Selected work
Representative Papers
A Field Guide to Decision Making
This study addresses the multifaceted challenges confronting decision-makers in high-stakes environments—namely, uncertainty, resource constraints, time pressure, and accountability risks. To navigate these complexities, the paper proposes an agent-based metadata governance mechanism that synergistically integrates machine intelligence with human cognition. By dynamically managing contextual metadata, this approach enhances situational awareness, establishes an adaptive decision-making framework, and balances risk tolerance with conditional accountability. Moving beyond conventional decision-support paradigms, the proposed method significantly improves contextual understanding, decision coherence, and adaptability in complex, time-critical scenarios, thereby offering a practical pathway toward responsible and effective decision-making in high-consequence settings.
0 citationsRead paper
BladeSDF : Unconditional and Conditional Generative Modeling of Representative Blade Geometries Using Signed Distance Functions
This work proposes an implicit generative framework based on DeepSDF to address the limitations of conventional turbine blade generation methods, which often lack performance awareness, manufacturability guarantees, and geometric continuity. By leveraging signed distance functions (SDFs), the approach achieves high-fidelity geometric reconstruction and constructs an interpretable, near-Gaussian latent space aligned with key aerodynamic and structural parameters, enabling both unconditional synthesis and performance-driven conditional generation. A compact neural network maps engineering performance metrics end-to-end to latent codes, facilitating efficient geometry generation through interpolation and Gaussian sampling. Experimental results demonstrate that the reconstructed surfaces exhibit distance errors within 1% of the blade’s maximum dimension, achieving high fidelity while maintaining strong generalization to unseen designs.
0 citationsRead paper
Runtime Energy Monitoring for RISC-V Soft-Cores
Real-time, low-overhead energy monitoring remains challenging for RISC-V soft-core processors during design space exploration, particularly due to reliance on complex microarchitectural models. Method: This paper proposes a hardware-assisted real-time energy monitoring approach that bypasses such models. It integrates an FPGA system-level module with a custom current/voltage measurement board to directly capture runtime electrical signals, exposing them via a memory-mapped interface for lightweight readout by a monitoring service—achieving high accuracy and low latency without consuming FPGA logic resources. Contribution/Results: The solution supports scalable deployment from single-node to multi-node clusters, enabling synchronized multi-point sampling and distributed analysis. Experimental evaluation demonstrates fine-grained energy-efficiency tracking during RISC-V soft-core execution of shallow neural networks. This provides empirical support for joint performance–energy optimization in power-constrained domains such as aerospace systems.
0 citationsRead paper
Recent publications
Latest Papers
The record is part of the task: matched-record evaluation of text classifiers across maintenance, safety and recall reporting
研究通过比较同一案例在不同阶段的记录，评估文本分类器在维护、安全和召回报告中的表现差异。
0 citationsRead paper
A Field Guide to Decision Making
This study addresses the multifaceted challenges confronting decision-makers in high-stakes environments—namely, uncertainty, resource constraints, time pressure, and accountability risks. To navigate these complexities, the paper proposes an agent-based metadata governance mechanism that synergistically integrates machine intelligence with human cognition. By dynamically managing contextual metadata, this approach enhances situational awareness, establishes an adaptive decision-making framework, and balances risk tolerance with conditional accountability. Moving beyond conventional decision-support paradigms, the proposed method significantly improves contextual understanding, decision coherence, and adaptability in complex, time-critical scenarios, thereby offering a practical pathway toward responsible and effective decision-making in high-consequence settings.
0 citationsRead paper
BladeSDF : Unconditional and Conditional Generative Modeling of Representative Blade Geometries Using Signed Distance Functions
This work proposes an implicit generative framework based on DeepSDF to address the limitations of conventional turbine blade generation methods, which often lack performance awareness, manufacturability guarantees, and geometric continuity. By leveraging signed distance functions (SDFs), the approach achieves high-fidelity geometric reconstruction and constructs an interpretable, near-Gaussian latent space aligned with key aerodynamic and structural parameters, enabling both unconditional synthesis and performance-driven conditional generation. A compact neural network maps engineering performance metrics end-to-end to latent codes, facilitating efficient geometry generation through interpolation and Gaussian sampling. Experimental results demonstrate that the reconstructed surfaces exhibit distance errors within 1% of the blade’s maximum dimension, achieving high fidelity while maintaining strong generalization to unseen designs.
0 citationsRead paper
Runtime Energy Monitoring for RISC-V Soft-Cores
Real-time, low-overhead energy monitoring remains challenging for RISC-V soft-core processors during design space exploration, particularly due to reliance on complex microarchitectural models. Method: This paper proposes a hardware-assisted real-time energy monitoring approach that bypasses such models. It integrates an FPGA system-level module with a custom current/voltage measurement board to directly capture runtime electrical signals, exposing them via a memory-mapped interface for lightweight readout by a monitoring service—achieving high accuracy and low latency without consuming FPGA logic resources. Contribution/Results: The solution supports scalable deployment from single-node to multi-node clusters, enabling synchronized multi-point sampling and distributed analysis. Experimental evaluation demonstrates fine-grained energy-efficiency tracking during RISC-V soft-core execution of shallow neural networks. This provides empirical support for joint performance–energy optimization in power-constrained domains such as aerospace systems.
0 citationsRead paper