Yufan Xia · Vol. III — Publications — Melbourne · MMXXVI
Tab. I

Selected publications

A short, hand-curated list. The full record is on Google Scholar; this page is meant for reading.

Filter
[ 01 ]

NN-xTB: density functional accuracy at semi-empirical speed with neural-network extended tight-binding

Yufan Xia, Albert Thie, Joshua Soon, Giuseppe Maria Junior Barca
Nature Communications— ★ Editors' Highlight
NN-xTB augments GFN2-xTB with neural-network-predicted parameter shifts, achieving DFT-like accuracy (WTMAD-2 of 4 kcal/mol on GMTKN55) at near semi-empirical cost while preserving physical interpretability and self-consistency. Selected for the Nature Communications Editors' Highlights in inorganic & physical chemistry.
2026 journal
[ 02 ]

Uncertainty-Gated Escalation for a Solver-in-the-Loop Molecular Surrogate

Yufan Xia
AI for Science Workshop, NeurIPS— Accepted
Accepted to the AI for Science Workshop at NeurIPS 2026 (Sydney, 12 December 2026).
2026 workshop
[ 03 ]

Machine-Learning-Driven Runtime Optimization of BLAS Level 3 on Modern Multi-Core Systems

Yufan Xia, Giuseppe Maria Junior Barca
IEEE IPDPS— Best Poster Award
ADSALA uses ML to predict optimal thread counts across all BLAS Level 3 operations, delivering 1.5–3.0× speedups on Intel and AMD multi-core systems.
2024 conference
[ 04 ]

A Machine-Learning Approach Towards Runtime Optimisation of Matrix Multiplication

Yufan Xia, Marco De La Pierre, Amanda S Barnard, Giuseppe Maria Junior Barca
IEEE IPDPS
Proof-of-concept for the ADSALA library: on-the-fly ML-based thread selection for GEMM, validated on Intel Cascade Lake and AMD architectures.
2023 conference