Add people_search benchmark (scorer-based, provider-agnostic) - #12

Open

diya-kejriwal wants to merge 8 commits into

youdotcom-oss:mainfrom

diya-kejriwal:people-search-framework

Commits on Jul 23, 2026

Add scorer-based people_search benchmark to the shared eval runner.

diya-kejriwal
and
cursoragent
committed
0e94383
Make people_search provider-agnostic via a generic HTTP sampler.

diya-kejriwal
and
cursoragent
committed
ead3b2b
Add overall and persona LLM judges to people_search grading.

diya-kejriwal
and
cursoragent
committed
5486291
Document people_search dataset, scorers, judges, and HTTP contract.

diya-kejriwal
and
cursoragent
committed
5035874

Commits on Jul 24, 2026

Clarify people_search metrics: do not call has-people “accuracy”.

diya-kejriwal
and
cursoragent
committed
5e00bfc
Align people_search README with the implemented eval behavior.

diya-kejriwal
and
cursoragent
committed
7dd6602
Wording: people_search works with any people-search HTTP endpoint.

diya-kejriwal
and
cursoragent
committed
d42e024
Leave people_search CSV answer column empty (no sample answers).

diya-kejriwal
and
cursoragent
committed
6a08c75