Skip to content

Document z_score normalization and rrf combination in Hybrid Optimizer #12483

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
RexyZhang2001 wants to merge 2 commits into
base: main
Choose a base branch
from
+9 −1
Open

Document z_score normalization and rrf combination in Hybrid Optimizer #12483

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
560ced2
Document z_score normalization and rrf combination in Hybrid Optimizer
RexyZhang2001 May 28, 2026
7da19f6
Link to upstream processor docs and clarify equal weights for RRF
RexyZhang2001 May 29, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 9 additions & 1 deletion _search-plugins/search-relevance/optimize-hybrid-search.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,10 +81,18 @@ PUT _plugins/_search_relevance/experiments

The hybrid search optimization experiment runs different evaluations based on the search configuration. The following parameters and parameter values are taken into account:

* Two normalization techniques: `l2` and `min_max`.
**Score-based variants**:

* Three normalization techniques: `l2`, `min_max`, and `z_score`.
* Three combination techniques: `arithmetic_mean`, `harmonic_mean`, `geometric_mean`.
* The lexical and neural search weights, which are values ranging from `0.0` to `1.0` in 0.1 increments.

Note that `z_score` is paired only with `arithmetic_mean` due to a [normalization-processor](https://docs.opensearch.org/latest/search-plugins/search-pipelines/normalization-processor/#request-body-fields) restriction.

**Rank-based variants**:

* The `rrf` ([Reciprocal Rank Fusion](https://docs.opensearch.org/latest/search-plugins/search-pipelines/score-ranker-processor/)) combination technique, evaluated with `rank_constant` values of `1`, `5`, `10`, `20`, and `60`. RRF variants use the default equal weights across sub-queries.

Every query in the query set is executed for all different parameter combinations, and the results are evaluated by using the judgment list.

## Evaluating the results
Expand Down