Surprise Calibration for Better In-Context Learning
This addresses performance degradation in dynamic in-context learning settings for NLP researchers and practitioners, offering an incremental improvement over prior bias calibration techniques.
The paper tackles biases in in-context learning for large language models by introducing Surprise Calibration, which uses surprise signals to adaptively adjust class priors, showing superior performance over existing methods on benchmark NLP tasks.
In-context learning (ICL) has emerged as a powerful paradigm for task adaptation in large language models (LLMs), where models infer underlying task structures from a few demonstrations. However, ICL remains susceptible to biases that arise from prior knowledge and contextual demonstrations, which can degrade the performance of LLMs. Existing bias calibration methods typically apply fixed class priors across all inputs, limiting their efficacy in dynamic ICL settings where the context for each query differs. To address these limitations, we adopt implicit sequential Bayesian inference as a framework for interpreting ICL, identify "surprise" as an informative signal for class prior shift, and introduce a novel method--Surprise Calibration (SC). SC leverages the notion of surprise to capture the temporal dynamics of class priors, providing a more adaptive and computationally efficient solution for in-context learning. We empirically demonstrate the superiority of SC over existing bias calibration techniques across a range of benchmark natural language processing tasks.