IBM / unitxt / 17430157886

81%

push

github

Improved multi turn evaluation to be self contained and use LLM as judge (#1929)

* Improved multi turn evaluation to be self contained and use LLM as judge

Also improved printout summary of instance score

* Moved to normalized sacrebleu

* Changed to clear use of typing.

Signed-off-by: Yoav Katz <katz@il.ibm.com>

* Added link to multi tuen concersation in the main examples page

Signed-off-by: Yoav Katz <katz@il.ibm.com>

* Removed unused prints

---------

Signed-off-by: Yoav Katz <katz@il.ibm.com>
Co-authored-by: Elron Bandel <elronbandel@gmail.com>

1595 of 1992 branches covered (80.07%)

Branch coverage included in aggregate %.

10839 of 13396 relevant lines covered (80.91%)

0.81 hits per line