LogitProd fuses logits from heterogeneous pathology foundation models via sample-adaptive weights, ranking first on 20 of 22 benchmarks with a 3% average gain over the best single model and 12x lower training cost than feature fusion.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Plug-and-Play Logit Fusion for Heterogeneous Pathology Foundation Models
LogitProd fuses logits from heterogeneous pathology foundation models via sample-adaptive weights, ranking first on 20 of 22 benchmarks with a 3% average gain over the best single model and 12x lower training cost than feature fusion.