LLM tasks are supported by multiple distinct circuits rather than unique mechanisms, demonstrated via Overlap-Aware Sheaf Repulsion and the Distributive Dense Circuit Hypothesis.
Relational Graph Attention Networks
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
abstract
We investigate Relational Graph Attention Networks, a class of models that extends non-relational graph attention mechanisms to incorporate relational information, opening up these methods to a wider variety of problems. A thorough evaluation of these models is performed, and comparisons are made against established benchmarks. To provide a meaningful comparison, we retrain Relational Graph Convolutional Networks, the spectral counterpart of Relational Graph Attention Networks, and evaluate them under the same conditions. We find that Relational Graph Attention Networks perform worse than anticipated, although some configurations are marginally beneficial for modelling molecular properties. We provide insights as to why this may be, and suggest both modifications to evaluation strategies, as well as directions to investigate for future work.
representative citing papers
GAT uses static attention where neighbor rankings ignore the query node and thus cannot express some graph problems; GATv2 enables dynamic attention and outperforms GAT on 11 OGB and other benchmarks with equal parameters.
AFGNN detects API misuses in Java code more effectively than prior methods by representing usage as graphs and clustering learned embeddings from self-supervised training.
LUMINA derives three design principles for physics-informed foundation models that balance accuracy, constraint satisfaction, and reliability on topology-transferable ACOPF problems.
The survey groups attention-based GNNs into three stages—graph recurrent attention networks, graph attention networks, and graph transformers—while reviewing architectures and future directions.
FOCAL fuses unconstrained coverage attention and meta-path anchoring attention to improve multi-label classification on heterogeneous graphs by resolving semantic dilution versus coverage constraint trade-offs.
BIC-Hunter combines confident learning for label denoising and GCNs on homogeneous graphs to identify bug-inducing commits, reporting gains of 6.16-7.13% on Recall@K and 8.43-32.82% on MFR over prior methods.
GA-S2S integrates T5 with RGAT to jointly process text and k-hop subgraph topology for knowledge graph link prediction, reporting up to 19% relative accuracy gain over seq2seq baselines on CoDEx.
citing papers explorer
-
All Circuits Lead to Rome: Rethinking Functional Anisotropy in Circuit and Sheaf Discovery for LLMs
LLM tasks are supported by multiple distinct circuits rather than unique mechanisms, demonstrated via Overlap-Aware Sheaf Repulsion and the Distributive Dense Circuit Hypothesis.
-
How Attentive are Graph Attention Networks?
GAT uses static attention where neighbor rankings ignore the query node and thus cannot express some graph problems; GATv2 enables dynamic attention and outperforms GAT on 11 OGB and other benchmarks with equal parameters.
-
AFGNN: API Misuse Detection using Graph Neural Networks and Clustering
AFGNN detects API misuses in Java code more effectively than prior methods by representing usage as graphs and clustering learned embeddings from self-supervised training.
-
LUMINA: Foundation Models for Topology Transferable ACOPF
LUMINA derives three design principles for physics-informed foundation models that balance accuracy, constraint satisfaction, and reliability on topology-transferable ACOPF problems.
-
Attention-based graph neural networks: a survey
The survey groups attention-based GNNs into three stages—graph recurrent attention networks, graph attention networks, and graph transformers—while reviewing architectures and future directions.
-
FOCAL-Attention for Heterogeneous Multi-Label Prediction
FOCAL fuses unconstrained coverage attention and meta-path anchoring attention to improve multi-label classification on heterogeneous graphs by resolving semantic dilution versus coverage constraint trade-offs.
-
Confident Learning-based Network for Detecting Bug-Inducing Commits on SZZ with Noisy Labels
BIC-Hunter combines confident learning for label denoising and GCNs on homogeneous graphs to identify bug-inducing commits, reporting gains of 6.16-7.13% on Recall@K and 8.43-32.82% on MFR over prior methods.
-
Leveraging Graph Structure in Seq2Seq Models for Knowledge Graph Link Prediction
GA-S2S integrates T5 with RGAT to jointly process text and k-hop subgraph topology for knowledge graph link prediction, reporting up to 19% relative accuracy gain over seq2seq baselines on CoDEx.