MLPs used as feature-transformation modules improve edge probing performance as much as or more than GNNs, while message-passing alone often degrades it.
Do PLMs know and understand ontological knowledge?
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Do We Really Need GNNs with Explicit Structural Modeling? MLPs Suffice for Language Model Representations
MLPs used as feature-transformation modules improve edge probing performance as much as or more than GNNs, while message-passing alone often degrades it.