Pith. sign in

REVIEW 5 cited by

Less is More: on the Over-Globalizing Problem in Graph 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 2405.01102 v2 pith:JMFTLPPH submitted 2024-05-02 cs.LG cs.AI

classification cs.LGcs.AI
keywords graphnodesattentionglobalinformationmechanismover-globalizingproblem
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Graph Transformer, due to its global attention mechanism, has emerged as a new tool in dealing with graph-structured data. It is well recognized that the global attention mechanism considers a wider receptive field in a fully connected graph, leading many to believe that useful information can be extracted from all the nodes. In this paper, we challenge this belief: does the globalizing property always benefit Graph Transformers? We reveal the over-globalizing problem in Graph Transformer by presenting both empirical evidence and theoretical analysis, i.e., the current attention mechanism overly focuses on those distant nodes, while the near nodes, which actually contain most of the useful information, are relatively weakened. Then we propose a novel Bi-Level Global Graph Transformer with Collaborative Training (CoBFormer), including the inter-cluster and intra-cluster Transformers, to prevent the over-globalizing problem while keeping the ability to extract valuable information from distant nodes. Moreover, the collaborative training is proposed to improve the model's generalization ability with a theoretical guarantee. Extensive experiments on various graphs well validate the effectiveness of our proposed CoBFormer.

Discussion (0). Sign in to comment.

Forward citations

Cited by 5 Pith papers

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

  1. Crys-JEPA: Accelerating Crystal Discovery via Embedding Screening and Generative Refinement

    cs.LG 2026-05 unverdicted novelty 6.0 of 10

    Crys-JEPA introduces a joint embedding predictive architecture that creates an energy-aware latent space, enabling embedding-based stability screening and a refinement pipeline that yields up to 72.7% gains on the V.S...

  2. Distance-Misaligned Training in Graph Transformers and Adaptive Graph-Aware Control

    cs.LG 2026-04 unverdicted novelty 6.0 of 10

    A controllable synthetic benchmark on contextual SBM graphs reveals distance-misaligned training in Graph Transformers, with an oracle adaptive controller improving performance by matching task-specific distance targets.

  3. Composable Crystals: Controllable Materials Discovery via Concept Learning

    cs.LG 2026-05 unverdicted novelty 5.0 of 10

    VQ-VAE concept learning enables controllable recombination of crystal motifs to generate structures with reported gains in validity-stability-uniqueness-novelty metrics on MP-20 and Alex-MP-20.

  4. Computational Control of Nonlinear Partial Differential Equations Using Machine Learning

    math.OC 2026-04 unverdicted novelty 5.0 of 10

    A physics-informed neural network method is developed to approximate controls for nonlinear PDEs, including convergence analysis and numerical experiments demonstrating good performance.

  5. Computational Control of Nonlinear Partial Differential Equations Using Machine Learning

    math.OC 2026-04 conditional novelty 4.0 of 10

    On synthetic CSBM graphs, a graph-distance bias knob systematically shifts Graph Transformers between over-globalizing and under-reaching regimes, and an oracle target-gap controller tracks the best fixed bias across ...

Pith tools