Selected work

Representative Papers

First- and Second-Order Phase Transformation Modeling Based on the Hamilton Principle: A Coupled Thermo-Mechanical Approach for Glass Additive Manufacturing

Jul 29, 2026

This study addresses the complex microstructural evolution and residual stress challenges in glass additive manufacturing induced by extreme thermal histories. The authors propose a unified variational framework based on the extended Hamilton’s principle, which for the first time couples thermo-mechanical and phase transformation processes to simultaneously capture both the first-order melting and second-order glass transition under large deformations. A kinetic freezing mechanism is incorporated to model glass formation. The framework integrates a temperature-dependent viscosity constitutive law, the single-slice neighborhood element method (NEM), and a three-dimensional finite element implementation in ANSYS. The approach successfully reproduces time–temperature–transformation (TTT) behavior across varying cooling rates and accurately predicts residual stresses and macroscopic warpage during laser-based deposition, establishing a high-fidelity multiphysics simulation foundation for glass additive manufacturing.

0 citationsRead paper

Leveraging AutoML for Sustainable Deep Learning: A Multi-Objective HPO Approach on Deep Shift Neural Networks

Jun 22, 2026

This work addresses the substantial environmental burden imposed by the high computational cost of deep learning models by introducing, for the first time, a multi-objective AutoML framework tailored for deep shift neural networks (DSNNs)—an energy-efficient alternative that has lacked systematic optimization. The proposed approach integrates multi-fidelity Bayesian optimization with mixed-precision quantization to automatically discover Pareto-optimal trade-offs between accuracy and energy consumption in image classification tasks. Notably, the study uncovers counterintuitive yet highly efficient low-precision configurations, achieving approximately 20% performance improvement and reducing carbon emissions by over 60%. The method’s effectiveness and generalizability are further validated across multiple backbone architectures, demonstrating its broad applicability in sustainable deep learning design.

0 citationsRead paper

Average Attention Transformers and Arithmetic Circuits

May 06, 2026

We analyse the computational power of transformer encoders as sequence-to-sequence functions on vectors. We show that average hard attention can be used to simulate arithmetic circuits if they are given as an input to an encoder. The circuit families that can be simulated this way have constant depth while using unbounded addition, binary multiplication and sign gates. The transformers we use have arithmetic circuits instead of feed-forward networks. With typical average attention the functions they compute are also computed by the same class of circuit families. Our results hold for transformers over the reals, rationals and any ring in between the two.

0 citationsRead paper

SWARM-SLR AIssistant: A Unified Framework for Scalable Systematic Literature Review Automation

Mar 05, 2026

This study addresses critical bottlenecks in current systematic literature review (SLR) tools, particularly their limited scalability and poor integration into user-friendly workflows. To overcome these challenges, the authors propose a unified framework that combines structured SLR methodologies with large language model (LLM)-based intelligent agents. The framework features modular interfaces for integrating diverse research tools and introduces a centralized, metadata-driven tool registry that enables developers to annotate and share tool specifications autonomously. This design substantially enhances the modularity, extensibility, and automation of SLR processes. Preliminary evaluation indicates marked improvements in system usability; however, balancing efficiency, accessibility, and transparency remains a significant challenge.

0 citationsRead paper

Aerospace.Wikibase: Towards a Knowledge Infrastructure for Aerospace Engineering

Mar 05, 2026

This work addresses the long-standing absence of a unified knowledge management infrastructure in aerospace engineering, which has led to data fragmentation and redundant research efforts. To remedy this, the study introduces Wikibase into the domain for the first time, establishing an open, extensible, and decentralized collaborative knowledge graph platform. Through a systematic literature review, the authors structurally extracted and integrated over 700 core terms, constructing a standardized, reusable, and interlinkable conceptual framework. The platform is designed to simultaneously support public knowledge co-creation and safeguard project-specific private information, thereby providing a sustainable infrastructural foundation for cross-team collaboration and long-term knowledge accumulation in aerospace engineering.

0 citationsRead paper

Recent publications

Latest Papers

First- and Second-Order Phase Transformation Modeling Based on the Hamilton Principle: A Coupled Thermo-Mechanical Approach for Glass Additive Manufacturing

Jul 29, 2026

This study addresses the complex microstructural evolution and residual stress challenges in glass additive manufacturing induced by extreme thermal histories. The authors propose a unified variational framework based on the extended Hamilton’s principle, which for the first time couples thermo-mechanical and phase transformation processes to simultaneously capture both the first-order melting and second-order glass transition under large deformations. A kinetic freezing mechanism is incorporated to model glass formation. The framework integrates a temperature-dependent viscosity constitutive law, the single-slice neighborhood element method (NEM), and a three-dimensional finite element implementation in ANSYS. The approach successfully reproduces time–temperature–transformation (TTT) behavior across varying cooling rates and accurately predicts residual stresses and macroscopic warpage during laser-based deposition, establishing a high-fidelity multiphysics simulation foundation for glass additive manufacturing.

0 citationsRead paper

Leveraging AutoML for Sustainable Deep Learning: A Multi-Objective HPO Approach on Deep Shift Neural Networks

Jun 22, 2026

This work addresses the substantial environmental burden imposed by the high computational cost of deep learning models by introducing, for the first time, a multi-objective AutoML framework tailored for deep shift neural networks (DSNNs)—an energy-efficient alternative that has lacked systematic optimization. The proposed approach integrates multi-fidelity Bayesian optimization with mixed-precision quantization to automatically discover Pareto-optimal trade-offs between accuracy and energy consumption in image classification tasks. Notably, the study uncovers counterintuitive yet highly efficient low-precision configurations, achieving approximately 20% performance improvement and reducing carbon emissions by over 60%. The method’s effectiveness and generalizability are further validated across multiple backbone architectures, demonstrating its broad applicability in sustainable deep learning design.

0 citationsRead paper

Average Attention Transformers and Arithmetic Circuits

May 06, 2026

We analyse the computational power of transformer encoders as sequence-to-sequence functions on vectors. We show that average hard attention can be used to simulate arithmetic circuits if they are given as an input to an encoder. The circuit families that can be simulated this way have constant depth while using unbounded addition, binary multiplication and sign gates. The transformers we use have arithmetic circuits instead of feed-forward networks. With typical average attention the functions they compute are also computed by the same class of circuit families. Our results hold for transformers over the reals, rationals and any ring in between the two.

0 citationsRead paper

SWARM-SLR AIssistant: A Unified Framework for Scalable Systematic Literature Review Automation

Mar 05, 2026

This study addresses critical bottlenecks in current systematic literature review (SLR) tools, particularly their limited scalability and poor integration into user-friendly workflows. To overcome these challenges, the authors propose a unified framework that combines structured SLR methodologies with large language model (LLM)-based intelligent agents. The framework features modular interfaces for integrating diverse research tools and introduces a centralized, metadata-driven tool registry that enables developers to annotate and share tool specifications autonomously. This design substantially enhances the modularity, extensibility, and automation of SLR processes. Preliminary evaluation indicates marked improvements in system usability; however, balancing efficiency, accessibility, and transparency remains a significant challenge.

0 citationsRead paper

Aerospace.Wikibase: Towards a Knowledge Infrastructure for Aerospace Engineering

Mar 05, 2026

This work addresses the long-standing absence of a unified knowledge management infrastructure in aerospace engineering, which has led to data fragmentation and redundant research efforts. To remedy this, the study introduces Wikibase into the domain for the first time, establishing an open, extensible, and decentralized collaborative knowledge graph platform. Through a systematic literature review, the authors structurally extracted and integrated over 700 core terms, constructing a standardized, reusable, and interlinkable conceptual framework. The platform is designed to simultaneously support public knowledge co-creation and safeguard project-specific private information, thereby providing a sustainable infrastructural foundation for cross-team collaboration and long-term knowledge accumulation in aerospace engineering.

0 citationsRead paper