Documentation for Chain of Thought in GEval
prescod opened this issue · comments
Is your feature request related to a problem? Please describe.
When I read the documentation about GEval, it is incomplete.
The example code shows a criteria
and evaluation_steps
in the same flow. Furthermore, it shows
evaluation_steps
with a single item, which is not very logical. If you had only one step, you would
use a criteria.
Describe the solution you'd like
I would like to see a realistic example of how evaluation_steps is meant to be used.
@prescod I see what you mean, we didn't explain how the execution flow works for different scenarios of providing evaluation steps and criteria.. thanks for the feedback we'll include more description and examples!
@prescod fixed: https://docs.confident-ai.com/docs/metrics-llm-evals thanks!