A network generalizes to unseen combinations exactly when its computation graph matches a graph that already solves the test set, its internal codes are unambiguous, and the codes carry no extra information.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
A Theoretical Analysis of Compositional Generalization in Neural Networks: A Necessary and Sufficient Condition
A network generalizes to unseen combinations exactly when its computation graph matches a graph that already solves the test set, its internal codes are unambiguous, and the codes carry no extra information.