mlsanityData to 7 Oct 2026

Escaping Oversquashing: Addressable and Support-Aware Global Memory for Message Passing Networks

arXiv 2608.02709 ↗First posted 3 Aug 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Virtual nodes are a natural tool against oversquashing: they replace long message-passing paths by a two-hop global route. But when many nodes share one global state, that shortcut can become a bottleneck itself. We study two properties of this global memory. First, addressability: under constant-margin address codes and a nonlinearity that amplifies this margin, multiplicative write/read maps provide M selectable memory rows with only O(log M) address-code dimensions. Cross-attention slots and a constrained ELU+1 bilinear memory both satisfy these conditions. Second, support awareness: normalized cross-attention has no self-key for a latent query to use as a reference. A learned private anchor supplies this reference, keeps the read bounded, and exposes the strength of the matching source mass. We demonstrate the merits of such properties on several instances of Two-Radius and Tree-NeighborsMatch: both addressable realizations solve the controlled tasks through depth 5, where pooled VNs of comparable or larger size reach about 10.6%.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 21 of its 27 references that are in the map. First posted on arXiv 3 Aug 2026.

Builds on

21 of 27 references.

References · 21

  1. 2018How Powerful are Graph Neural Networks?cited by 4,355
  2. 2017Neural Message Passing for Quantum Chemistrycited by 3,460
  3. 2020Open Graph Benchmark: Datasets for Machine Learning on Graphscited by 2,148
  4. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
  5. 2020Transformers are RNNs: Fast Autoregressive Transformers with Linear Attentioncited by 1,979
  6. 2021Perceiver: General Perception with Iterative Attentioncited by 1,089
  7. 2018Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networkscited by 878
  8. 2023Vision Transformers Need Registerscited by 844
  9. 2020On the Bottleneck of Graph Neural Networks and its Practical Implicationscited by 624
  10. 2020Principal Neighbourhood Aggregation for Graph Netscited by 477
  11. 2021Linear Transformers Are Secretly Fast Weight Programmerscited by 471
  12. 2021Understanding over-squashing and bottlenecks on graphs via curvaturecited by 471
9 more
  1. 2022Long Range Graph Benchmarkcited by 258
  2. 2019What graph neural networks cannot learn: depth vs widthcited by 195
  3. 2023On Over-Squashing in Message Passing Neural Networks: The Impact of Width, Depth, and Topologycited by 163
  4. 2023Quantizable Transformers: Removing Outliers by Helping Attention Heads Do Nothingcited by 147
  5. 2023On the Connection Between MPNN and Graph Transformercited by 60
  6. 2024Distinguished In Uniform: Self Attention Vs. Virtual Nodescited by 19
  7. 2019Improving Attention Mechanism in Graph Neural Networks via Cardinality Preservationcited by 17
  8. 2024Understanding Virtual Nodes: Oversquashing and Node Heterogeneitycited by 14
  9. 2024NetDiff: Graph Diffusion with Improved Global Capabilities to Generate and Update Mobile Network Topologiescited by 2

Built on by

Not cited yet.