IBM / unitxt / 13830562050

81%

push

github

Safety eval updates (#1668)

* Remove redundant loop

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

* Add llamaguard metric to simple_safety_tests

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

* Updated llamaguard template and metric, using more accurate instructions.

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

* simple_safety_tests to use Granite Guardian harm metric.

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

* .

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

* both judges added

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

---------

Signed-off-by: Jonathan Bnayahu <bnayahu@il.ibm.com>

1554 of 1918 branches covered (81.02%)

Branch coverage included in aggregate %.

9722 of 12051 relevant lines covered (80.67%)

0.81 hits per line