Fixing a few bugs + use None when missing predictions

deepset-ai / haystack-evaluation

Using Haystack to benchmark different architectures over different datasets

Apache License 2.0

5 stars 0 forks source link

Closed mrm1001 closed 1 month ago

mrm1001 commented 1 month ago

If the LLM cannot predict and returns the string "None", we should treat this as None prediction, and ideally this should not affect the scores downstream.
Random sampling fixed
The csv for the scores missed the names of the metrics
Added the contexts to the csv file