StarFT improves the OOD robustness of fine-tuned zero-shot models by regularizing the output distribution on LLM-generated spurious textual descriptions to match the frozen zero-shot model.
Robust fine-tuning of zero-shot models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
StarFT: Robust Fine-tuning of Zero-shot Models via Spuriosity Alignment
StarFT improves the OOD robustness of fine-tuned zero-shot models by regularizing the output distribution on LLM-generated spurious textual descriptions to match the frozen zero-shot model.