On the Relation between Sensitivity and Accuracy in In-Context Learning

Anonymous

Keywords: in-context learning, few-shot learning, sensitivity, selective prediction, abstention, calibration, prompting

Abstract: In-context learning (ICL) suffers from oversensitivity to the prompt, making it unreliable in real-world scenarios. We study the sensitivity of ICL with respect to multiple perturbation types. First, we find that label bias obscures the true sensitivity, and therefore prior work may have significantly underestimated ICL sensitivity. Second, we observe a strong negative correlation between ICL sensitivity and accuracy: predictions sensitive to perturbations are less likely to be correct. Motivated by these findings, we propose \textsc{SenSel}, a few-shot selective prediction method that abstains from sensitive predictions. Experiments on ten classification datasets show that \textsc{SenSel} consistently outperforms a commonly used confidence-based baseline on abstention decisions.

Paper Type: short

Research Area: Machine Learning for NLP