MAG selects in-context demonstrations by propagating relevance scores over text and image graphs, using a small set of pseudo-labeled unlabeled samples to boost few-shot multimodal accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MAG: MAnifold Guided Semi-Supervised Multi-modal In-Context Learning
MAG selects in-context demonstrations by propagating relevance scores over text and image graphs, using a small set of pseudo-labeled unlabeled samples to boost few-shot multimodal accuracy.