Pith. sign in

REVIEW 8 cited by

GraphiT: Encoding Graph Structure in Transformers

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2106.05667 v1 pith:DKDD5SBS submitted 2021-06-10 cs.LG

classification cs.LG
keywords encodinggraphgraphitgraphsinformationmodelpositionalwell
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We show that viewing graphs as sets of node features and incorporating structural and positional information into a transformer architecture is able to outperform representations learned with classical graph neural networks (GNNs). Our model, GraphiT, encodes such information by (i) leveraging relative positional encoding strategies in self-attention scores based on positive definite kernels on graphs, and (ii) enumerating and encoding local sub-structures such as paths of short length. We thoroughly evaluate these two ideas on many classification and regression tasks, demonstrating the effectiveness of each of them independently, as well as their combination. In addition to performing well on standard benchmarks, our model also admits natural visualization mechanisms for interpreting graph motifs explaining the predictions, making it a potentially strong candidate for scientific applications where interpretation is important. Code available at https://github.com/inria-thoth/GraphiT.

Discussion (0). Sign in to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Logarithmic High-Probability Regret for Online Convex Optimization with Two-Point Bandit Feedback

    cs.LG 2026-03 unverdicted novelty 8.0 of 10

    Standard two-point projected gradient achieves fixed-comparator high-probability logarithmic regret for strongly convex OCO with two-point bandit feedback, with a leading d (not d²) horizon term.

  2. HydroGAT: Distributed Heterogeneous Graph Attention Transformer for Spatiotemporal Flood Prediction

    cs.LG 2025-09 conditional novelty 6.0 of 10

    A heterogeneous pixel-level graph attention transformer reaches NSE up to 0.97 for hourly flood prediction on two Midwest basins and scales to 64 GPUs.

  3. SaGIF: Improving Individual Fairness in Graph Neural Networks via Similarity Encoding

    cs.LG 2025-06 conditional novelty 6.0 of 10

    SaGIF adds an independent similarity encoder, initialized from a fused feature-and-topology oracle, to regular GNNs and reports better individual fairness on six benchmark graphs.

  4. Vector Representations of Vessel Trees

    eess.IV 2025-06 conditional novelty 6.0 of 10

    VeTTA encodes a vascular tree into one vector and recursively decodes it into a geometrically accurate, topologically valid tree, outperforming voxel-based autoencoders on reconstruction metrics.

  5. Graph Positional Autoencoders as Self-supervised Learners

    cs.LG 2025-05 conditional novelty 6.0 of 10

    A dual-path graph autoencoder that reconstructs node features and Laplacian-eigenvector distances reports strong self-supervised results on heterophilic and molecular benchmarks, with some overstatement in the margins...

  6. Simple Path Structural Encoding for Graph Transformers

    cs.LG 2025-02 conditional novelty 6.0 of 10

    SPSE replaces random walk edge encodings with simple path counts and outperforms RWSE on most of the tested graph benchmarks, especially for cycle-rich molecular graphs.

  7. GITO: Graph-Informed Transformer Operator for Learning Complex Partial Differential Equations

    cs.LG 2025-06 conditional novelty 5.0 of 10

    GITO, a graph-informed transformer operator, reports lower relative L2 errors than existing transformer-based neural operators on Navier-Stokes, heat conduction, and airfoil benchmark datasets.

  8. SpectralGap: Graph-Level Out-of-Distribution Detection via Laplacian Eigenvalue Gaps

    cs.LG 2025-05 conditional novelty 4.0 of 10

    SpecGap adjusts GNN features by subtracting the second-largest Laplacian eigenvector component times the spectral gap, claiming improved graph OOD detection without retraining.

Pith tools