The paper introduces gated graph attention and a learnable temperature to enhance standard graph attention networks. Gating filters out unreliable feature dimensions, while the temperature parameter adjusts the sharpness of attention coefficient distributions. Experiments on node classification and heterophilic benchmarks show that these modifications improve performance and robustness, with theoretical analysis supporting the observed gains.
By Zhongtian Ma, Hao Wu, Yexin Zhang, Qiaosheng Zhang, Chao Gao
arXiv:2610.28618v1 Announce Type: new
Abstract: Graph generative models commonly use graph-specific Transformer architectures to jointly update node and edge features. However, these architectures do...
By Vincent Jung, Alba Carballo-Castro, Yiming Qin, Lonneke van der Plas, Pascal Frossard
arXiv:2609.14663v1 Announce Type: new
Abstract: Deep neural networks are highly over-parameterized, and different parameter values represent the same predictive function. This makes their effective c...
By Vishnu Varadarajan, Mihir More, Aritra Das, Debayan Gupta
arXiv:2508. 17821v3 Announce Type: replace-cross Abstract: This paper investigates the limitations of the normalization in attention mechanisms.
By Timur Mudarisov, Mikhail Burtsev, Tatiana Petrova, Radu State
arXiv:2604. 07492v2 Announce Type: replace-cross Abstract: Message Passing Neural Networks have recently become the most popular approach to graph machine learning tasks; however, their receptive field is limited by the number of message passing layers.
By Oleg Platonov, Liudmila Prokhorenkova
FeatureFormer is a neural performance predictor that adds explicit node-wise encodings of FLOPs, parameter counts, and memory proxies to a gated graph attention architecture. It is designed to improve latency and energy prediction for neural networks on edge devices, addressing the limitation of existing GNN and transformer predictors that largely ignore node-level computational cost. The authors also introduce NNEQ, a large-scale energy consumption dataset, and show through extensive experiments that FeatureFormer achieves state‑of‑the‑art performance across both metrics, including challenging out‑of‑domain settings, while the encoding can broadly enhance existing predictors with negligible overhead.
By Matthew Grenier, William Hammer, Andrew Heuer, Nikhil Krishna, Yi Wang, Ramtin Zand
The paper shows that a plain Transformer can serve as an effective graph learner by adding three lightweight modifications: simplified L₂ attention, adaptive RMS normalization, and an MLP-based positional encoding stem. These changes preserve token magnitude and enable the model to achieve high expressivity on graph benchmarks, outperforming more complex graph transformer variants. The results suggest that plain Transformers can act as a unified backbone for multimodal learning across language, vision, and graph domains.
By Liheng Ma, Soumyasundar Pal, Yingxue Zhang, Philip H. S. Torr, Mark Coates
arXiv:2604. 17324v2 Announce Type: replace-cross Abstract: Global self-attention drives modern graph transformers, yet the softmax at its core imposes a structural constraint rarely examined directly: every attention row is non-negative and sums to one, so each per-head output is a mass-conserving convex combination of value vectors.
By Yang Liu, Dongxin Guo, Tom Zheng, Siu Ming Yiu, Liam Ning, Jikun Wu
arXiv:2509. 24770v2 Announce Type: replace Abstract: Large Language Models (LLMs) often generate incorrect or unsupported content, known as hallucinations.
By Fabrizio Frasca, Guy Bar-Shalom, Yftah Ziser, Haggai Maron
The paper introduces an asynchronous message‑passing framework for Graph Neural Networks to mitigate oversquashing, a problem where distant nodes cannot effectively communicate due to structural bottlenecks. Unlike conventional synchronous updates, the method updates a centrality‑guided batch of nodes at each layer, allowing information to propagate sequentially and reducing the need for increased channel capacity. Experiments on six standard and two long‑range graph classification benchmarks show notable performance gains, including 5 % improvement on REDDIT‑BINARY and 4 % on Peptides‑struct.
By Kushal Bose, Swagatam Das
arXiv:2607. 01553v1 Announce Type: cross Abstract: Transformers have become general-purpose architectures, but their all-to-all self-attention is poorly matched to graph data, whose interactions are sparse, structured and multi-scale.
By Leyan Li, Rennong Yang, Zhenxing Zhang, Liping Hu
The paper reinterprets graph neural networks (GNNs) as retrieval-augmented models, where each layer uses an MLP on a node representation and a permutation‑invariant summary of retrieved graph context instead of traditional message passing. It introduces RTA, a lightweight MLP‑based framework that replaces structural message passing with label‑aware retrieval and propagation, and provides theoretical links to softmax‑attention message passing and robustness to mis‑retrieved outliers. Experiments on text‑attributed graph benchmarks demonstrate that RTA matches or surpasses strong GNN and graph LLM baselines while improving efficiency and robustness.
By Jintang Li, Yuhong Chen, Ruofan Wu, Binli Luo, Jiayi Ji, Hui Li, Rongrong Ji