Repository navigation
MAINT: declare LatentInjection ingredients as required data - #3547
Merged
Roman Lutz (romanlutz) merged 36 commits intoNov 11, 2026
Merged
Roman Lutz (romanlutz) merged 36 commits into
Roman Lutz (romanlutz) merged 36 commits into
Conversation
…d-data # Conflicts: # tests/unit/scenario/garak/test_api_key.py
…ta-latent-injection
…s to selected techniques A required-data refresh committed before the scenario assembled its population, so a refresh that the assembly or the population validators would reject had already replaced the stored datasets, breaking resume of earlier runs. The refresh now runs the scenario's full unsampled assembly against the candidates first, so a rejected refresh changes nothing. Legacy dataset_config validators wrote against the raw template seeds, so they move onto the required data when an old configuration migrates, instead of receiving the assembled objectives. ApiKey now requires a prompt template only for the techniques a run selects.
…ta-latent-injection
… and still resumes
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
…work A validator can be any callable, including an unhashable instance such as a plain dataclass, so the migration can't collect them in a set. Remove the moved validators by object identity instead.
…ta-latent-injection
…on test The configuration no longer accepts its ingredients as dataset sources, so the population test assembles them through the required-data path instead of the old resolver.
…into feat/required-data-prompt-inject
…ta-latent-injection
…n' into feat/required-data-latent-injection
…ject empty population Required data and the population configuration are now fetched first and validated as one batch: when either half refreshes, each configuration and the scenario's full unsampled assembly run against all candidates before anything is written. A refresh of templates now sees records fetched in the same run, a valid joint refresh is accepted, and a records-only refresh that the assembly rejects leaves the stored rows and ids unchanged. A global fetch override now reaches required sources without their own policy even when the population has sources, and a source selection that names only required datasets is rejected instead of silently restoring the default population.
…ataset names A compound population names its datasets through its children, so its own sources are empty and its parent fetch is not the requested one. Required data now keeps its resolved fetch policy whenever the population names any dataset, so fetch=never and fetch=always reach the required sources for compound populations too. Preparing required data and the population together now rejects a dataset named by both before anything is fetched. Candidates are keyed by name, so a shared name validated one batch and then committed both.
… contract A compound configuration has no root sources even when its children name datasets, so it passed the source guard in select_assembled_groups and its children's validators and limits were skipped. It now raises a clear error instead; use a single source-free configuration. The docs said required data is read without filters, but its filters do narrow the input rows. The module, Scenario.get_required_data() and the scenario instructions now say every row that remains after filtering is read, without sampling or size caps.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
…tent-injection # Conflicts: # tests/unit/scenario/test_default_run_size_estimates.py
Utkarsh Bahuguna (u7k4rs6)
marked this pull request as ready for review
November 10, 2026 08:59
Roman Lutz (romanlutz)
approved these changes
Nov 11, 2026
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.
Third scenario for #3336, following PromptInject in #3546. The five LatentInjection ingredient datasets move to get_required_data, and the reserved boundary marker check becomes a required data validator so refreshed data gets checked too. LatentInjectionDatasetConfiguration keeps the families, max_total and the family/trigger coverage sampling, and its own uncapped and exact-datasets checks are gone. Filters and fetch settings on the config still reach the ingredients, and passing the ingredients as sources still works with a deprecation warning.
Part of #3336