Repository navigation
Refresh BCAL gold dataset criteria and remove unused manifest - #996
Merged
Merged
Conversation
Contributor
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
The changed pass criteria lack the required major benchmark version bump, allowing runs using old and corrected criteria to be aggregated together.
1 open issue
What changed in this PR
Refreshes BCAL gold evaluation criteria to accept valid alternative AL implementations, providing a corrected dataset baseline for the scenario-evidence and scoring work.
Changes:
- Updates assertion text for 4 of 110 gold entries without changing prompts or IDs.
- Removes the obsolete NL2AL manifest following the transition from panels to tiers.
| File | Description |
|---|---|
dataset/nl2al.jsonl |
Corrects criteria for notifications, Boolean defaults, complaint summaries, and aisle queries. |
dataset/nl2al_manifest.json |
Deletes unused dataset version, hash, and panel metadata. |
Review effort: Balanced
Sun Haoran (haoranpb)
marked this pull request as draft
November 2, 2026 16:02
Sun Haoran (haoranpb)
marked this pull request as ready for review
November 3, 2026 14:38
Esben Nyhuus Kristoffersen (esbenk)
approved these changes
Nov 9, 2026
Nothing reads dataset/nl2al_manifest.json since #974 replaced dataset panels and their hash/version checks with tiers. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Sun Haoran (haoranpb)
force-pushed
the
dataset/bcal-gold-criteria-fixes
branch
from
November 9, 2026 13:08
46dfa79 to
b25bc3f
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.

Lands the BCAL gold dataset correction that was stranded on
dataset/sync-bcal-2026-10-14(no PR), so it can serve as the base of the BCal evidence stack (#984 → #985, #992).Changes
9d444126, cherry-picked): evaluator assertion corrections synced from BC-DeveloperExperiencecandidate-v2. Onlyexpected[].textchanges, on 4 of 110 entries; prompts and IDs are unchanged. The criteria now accept valid alternative AL solutions:persona-item-low-stock-warning-consultant-2: drops the "message identifies the current item" requirementgl-account-needs-review-flag-2: an explicitInitValue = falseis also acceptedcollected-shipment-complaint-log-1: Item Card has no General group, so "after Item" counts as "after General";Method = Year/Month/Countquery columns are validexpanded-left-outer-slot-directory-1:TableRelationwith default validation satisfies the aisle check; bothOrderByforms are validdataset/nl2al_manifest.json: nothing has read it since Replace NL2AL dataset panels with a generic metadata tier filter #974 replaced panels and their hash/version checks with tiers.Dropped from the original branch
NL2ALDataset.versionbump and manifest SHA (removed by Replace NL2AL dataset panels with a generic metadata tier filter #974;benchmark_version+ commit SHA identify the dataset).cc98ace4(manifestsourceblock,docs/bcal-datasets.mdsection,test_nl2al_panel_selection.pyassertion): all three targets were deleted by Replace NL2AL dataset panels with a generic metadata tier filter #974.