arXiv:2611.03836 · v1

A Conditional Mutual Information Analysis of Response Variety in Language Models

首次提交 2026-11-03 03:10

版本更新 2026-11-03 03:10

雷达观测 2026-11-04 23:41

摘要

Large language models provide a variety of responses at inference (generation) that conventionally depend on a temperature parameter. Using a conditional mutual information approach, a quantitative, entropy based analysis of response variety is given. Three limiting behaviors are defined: arbiter, independent (conditionally) and neutral (deterministic) that span no to maximal mutual information. We give examples of each behavior and suggest their application to: temperature analysis, language analysis, prompt compression, and knowledge graphs.

作者备注：8 pages