Skip to content

Preserve input order when loading RAG results - #162

Merged
ofermend merged 1 commit into
vectara:devfrom
GioiaZheng:fix/deterministic-result-order
Oct 8, 2026
Merged

ofermend merged 1 commit into
vectara:devfrom
GioiaZheng:fix/deterministic-result-order

Conversation

@GioiaZheng

Copy link
Copy Markdown
Contributor

Summary

  • preserve the first-seen (query_id, query_run) order when loading RAG results
  • add a regression test covering non-lexicographically ordered query IDs

pandas.DataFrame.groupby sorts group keys by default. When query IDs are generated independently between runs, that sorting can change the relative order of queries in the evaluation output. Passing sort=False preserves the stable input order without changing grouping or evaluation semantics.

Closes #67.

Testing

  • pytest tests/test_rag_results_loader.py -q
  • git diff --check

@ofermend
ofermend requested review from vish119 and a balanced review from Copilot and removed request for vish119 October 8, 2026 02:17

@ofermend ofermend left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟢 Approval recommended

The focused ordering change has targeted regression coverage and no identified blocking issues.

0 open findings

What changed in this PR

Preserves input query order when loading RAG results, addressing the loader’s contribution to ordering instability reported in #67.

Changes:

  • Disables group-key sorting while retaining existing grouping behavior.
  • Adds a regression test using non-lexicographically ordered query IDs.
File Description
tests/​test_rag_results_loader.py Adds input-order regression coverage.
open_rag_eval/​rag_results_loader.py Uses sort=False to preserve first-seen group order.

🧠 Review effort: Balanced


Give feedback about Copilot approvals in this survey to enter a drawing for a $150 gift card.

@ofermend
ofermend merged commit e05341b into vectara:dev Oct 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Make ordering of evaluation results deterministic

4 participants