Large vision-language models score below 50 percent on a new cross-level compositional consistency benchmark, and a progressive meta-optimization training framework improves consistency on several VQA and TVG baselines.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Consistency of Compositional Generalization across Multiple Levels
Large vision-language models score below 50 percent on a new cross-level compositional consistency benchmark, and a progressive meta-optimization training framework improves consistency on several VQA and TVG baselines.