FEAT Add BeaverTails dataset loader

[romanlutz]

Add remote dataset loader for BeaverTails (PKU-Alignment/BeaverTails), containing 330k+ QA pairs annotated across 14 harm categories for safety alignment research. Filters to unsafe entries by default.

[Copilot]

Pull request overview

Adds a new remote seed dataset loader for the BeaverTails HuggingFace dataset, making it discoverable via SeedDatasetProvider and documenting its availability.

Changes:

• Introduces _BeaverTailsDataset remote loader with optional unsafe_only filtering (default: unsafe only).
• Registers the loader in the remote datasets module and adds unit tests for filtering behavior.
• Updates the “Loading Built-in Datasets” notebook output to include the new dataset name.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
pyrit/datasets/seed_datasets/remote/beaver_tails_dataset.py	New HuggingFace-backed loader that converts BeaverTails rows into SeedPrompts (unsafe-only by default).
pyrit/datasets/seed_datasets/remote/__init__.py	Imports/exports the new loader so it’s auto-registered/discoverable.
tests/unit/datasets/test_beaver_tails_dataset.py	Adds unit tests covering unsafe-only vs all-entries behavior and dataset naming.
doc/code/datasets/1_loading_datasets.ipynb	Notebook updated to reflect the new dataset in the available list (but now includes executed outputs/metadata).

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 5 comments.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

pyrit/prompt_converter/braille_converter.py:133

• In _get_braile, is_number is still reset to False after processing any character that isn’t in numberPunctuations (line 132). Since digits aren’t in numberPunctuations, this resets the number-mode after every digit, causing the Braille number indicator (characterUnicodes['num']) to be emitted before each digit instead of once per digit sequence. Consider resetting number-mode only when leaving a numeric run (e.g., when is_number is True and the current char is neither a digit nor an allowed number punctuation).

        is_number = False
        for char in text:
            if char in escapeCharacters:
                output += char
            elif char.isupper():
                if char.lower() in characterUnicodes:
                    output += characterUnicodes["caps"]
                    output += characterUnicodes[char.lower()]
            elif char in characterUnicodes:
                if char.isdigit() and not is_number:
                    is_number = True
                    output += characterUnicodes["num"]
                output += characterUnicodes[char]
            if is_number and char not in numberPunctuations:
                is_number = False

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

[varunj-msft]

Loader looks clean. Worth splitting the braille and markdown printer fixes into their own PRs? The braille one may need a follow up on the reset condition