Skip to main content

​
Overview

The geval diff command compares eval results between two runs to identify changes and regressions.

​
Usage

geval diff --previous <path> --current <path> [options]

​
Options

OptionDescription
-p, --previous <path>Path to previous eval results (required)
-c, --current <path>Path to current eval results (required)
--jsonOutput results as JSON
--no-colorDisable colored output

​
Example

geval diff --previous baseline.json --current new.json

​
Output

The diff command shows:
  • Metrics that improved
  • Metrics that regressed
  • Metrics that remained unchanged
  • Detailed comparison for each metric