Published in arXiv preprint arXiv:2610.10018, 2026
ONDA combines learned matrix-valued sheaf transport with second-order oscillatory dynamics for long-range graph learning. A stalk-wise sensitivity analysis characterises persistent influence in the conservative setting, while experiments demonstrate strong performance on long-range propagation, severe graph bottlenecks, and graph transfer, alongside competitive results on heterophilic benchmarks.
Published in arXiv preprint arXiv:2610.01494, 2026
A bridge between attention and sheaf neural networks: Top-A shows that multi-head attention is diagonal matrix-valued transport on its head space and adds edge-conditioned off-diagonal routes, letting each interaction move information across heads before aggregation. It recovers vanilla attention exactly when routing vanishes and improves relational, heterogeneous and algorithmic reasoning.
Published in arXiv preprint arXiv:2608.28853, 2026
ESNN learns directed, matrix-valued transport between neighbouring vector features while preserving exact Euclidean equivariance. It captures radial and tangential geometric interactions, supports controlled symmetry relaxation, and improves performance across dynamics, mesh simulation, point clouds, and molecular-property prediction.
Published in arXiv preprint arXiv:2605.10414, 2026
GAPE (Gated Adaptive Positional Encoding) addresses core limitations of RoPE in long-context language models. A content-aware bias is injected directly into attention logits while preserving rotary geometry: query-dependent and key-dependent gates suppress irrelevant distant tokens while protecting salient context, improving attention sharpness and long-context performance on retrieval and standard benchmarks.
Published in arXiv preprint arXiv:2512.00242, 2025
ArXiv preprint on Polynomial Neural Sheaf Diffusion: A Spectral Filtering Approach on Cellular Sheaves.
Published in Transactions on Graph Intelligence and Network Applications (TGINA), Scilight, 2026, 2026
A journal survey of graph-driven vulnerability intelligence: how software vulnerability detection, classification, localization and repair become graph learning problems over program entities (ASTs, CFGs, DFGs, PDGs, CPGs). The survey covers graph-based, multimodal and LLM-assisted methods, the dataset/tool/framework ecosystem, and five open challenges — from label noise to the new attack surfaces of AI-generated code and agents.
Published in NeurReps Workshop @ NeurIPS 2026 - Symmetry and Geometry in Neural Representations, (Sydney, Australia 🇦🇺), 2026
Accepted at the NeurReps Workshop @ NeurIPS 2026. HetSheaf is a cellular-sheaf framework for heterogeneous graphs that encodes node and edge types through type-aware local feature spaces and learned restriction maps, without specialised architectural components. The companion SheafPool readout is invariant to basis changes and enables graph-level prediction. Gains of up to +2 pp on the Heterogeneous Graph Benchmark with up to 10× fewer parameters.
Published in CVPR (Computer Vision and Pattern Recognition) 2025 Workshops (Nashville, USA 🇺🇸), 2025
Z-SASLM introduces a zero-shot, fine-tuning-free approach to style alignment in diffusion models by blending multiple reference styles directly in latent space using spherical linear interpolation (SLI) with learned, context-aware weights. The method avoids model retraining, preserves content semantics, and yields consistent style transfer across prompts and seeds.