COSMO fuses a source classifier and a vision-language model through an entropy-weighted consensus, then regulates consensus drift during source-free domain adaptation, achieving state-of-the-art accuracy on four benchmarks.
Machine Learning , volume=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
COSMO: Consensus-Driven Shift Modulation for Source-Free Domain Adaptation
COSMO fuses a source classifier and a vision-language model through an entropy-weighted consensus, then regulates consensus drift during source-free domain adaptation, achieving state-of-the-art accuracy on four benchmarks.