2024/02/18 by Zhichao Xu, Xu, Zhichao, Daniel Cohen +5 · 3 citations
Business, Management and Accounting · Engineering · #Business Process Modeling and Analysis #Computation and Language (cs.CL) #FOS: Computer and information sciences #Flexible and Reconfigurable Manufacturing Systems #Manufacturing Process and Optimization
paper · pdf · doi:10.48550/arxiv.2402.11447
openalex publication_date 2024/02/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
By allowing models to predict without task-specific training, in-context learning (ICL) with pretrained LLMs has enormous potential in NLP. However, a number of problems persist in ICL. In particular, its performance is sensitive to the choice and order of in-context examples. Given the same set of in-context examples with different orderings, model performance may vary between near random to near state-of-the-art. In this work, we formulate in-context example ordering as an optimization problem. We examine three problem settings that differ in the assumptions they make about what is known about the task. Inspired by the idea of learning from label proportions, we propose two principles for in-context example ordering guided by model's probability predictions. We apply our proposed principles to thirteen text classification datasets and nine different autoregressive LLMs with 700M to 13B parameters. We demonstrate that our approach outperforms the baselines by improving the classification accuracy, reducing model miscalibration, and also by selecting better in-context examples.