A contrastive alignment objective applied during few-shot fine-tuning improves cross-lingual transfer for NLI and relation extraction in many settings, though the reported advantage over prompting baselines is inconsistent across models and datasets.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Bridging Language Gaps: Enhancing Few-Shot Language Adaptation
A contrastive alignment objective applied during few-shot fine-tuning improves cross-lingual transfer for NLI and relation extraction in many settings, though the reported advantage over prompting baselines is inconsistent across models and datasets.