PromptNCE: Pointwise Mutual Information Predictions Using Only LLMs and Contrastive Estimation Prompts

This paper introduces PromptNCE, a method to estimate pointwise mutual information (PMI) using only LLMs and contrastive estimation prompts, circumventing the need for task-specific critics. It presents a benchmark with human-derived PMI and shows PromptNCE achieves Spearman correlation up to 0.82.