CIPE constructs graph positional encodings from communicability so that self-attention similarities equal the sum of all-path contributions between nodes, yielding 35.5% average gains on seven benchmarks over structure-agnostic Transformers.
Graphit: Encoding graph structure in transformers,
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.LG 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
Random Add-Drop Edge (RADE) jointly drops and adds edges stochastically during GNN training, with provable train-inference alignment and an adaptive rate balancer, to regularize against overfitting and mitigate over-squashing.
citing papers explorer
-
Communicability-Inspired Positional Encoding (CIPE)
CIPE constructs graph positional encodings from communicability so that self-attention similarities equal the sum of all-path contributions between nodes, yielding 35.5% average gains on seven benchmarks over structure-agnostic Transformers.
-
RADE: Random Add-Drop Edge as a Regularizer
Random Add-Drop Edge (RADE) jointly drops and adds edges stochastically during GNN training, with provable train-inference alignment and an adaptive rate balancer, to regularize against overfitting and mitigate over-squashing.