Home

EVA is a response evaluation tool specifically designed to assess outputs from large language models (LLMs). By classifying these responses as either affirmative or similar, EVA plays a critical role in the ongoing testing and development of LLMs. This wiki provides detailed insights and guidance on how to utilize EVA effectively as part of the Trust4AI research project.

Table of contents

1. Home
2. User Guide
   1. Evaluation Methods
      1. Yes/no Evaluation
      2. Three Reasons Evaluation
      3. WH Question Evaluation
      4. Multiple Choice Evaluation