TEST: Add unit tests for remote dataset loaders

[romanlutz]

Summary

Adds unit tests for 20 previously untested remote dataset loader files.

New Test Files

Each test mocks HTTP/HuggingFace calls and tests dataset fetching, filtering, and validation:

aegis_ai_content_safety, aya_redteaming, ccp_sensitive_prompts, darkbench,
equitymedqa, forbidden_questions, harmbench, harmbench_multimodal,
jbb_behaviors, librai_do_not_answer, llm_latent_adversarial_training,
medsafetybench, mlcommons_ailuminate, multilingual_vulnerability,
pku_safe_rlhf, red_team_social_bias, sorry_bench, sosbench,
tdc23_redteaming, xstest

Testing

All 69 tests pass locally.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds pytest unit coverage for previously untested remote seed dataset loaders by mocking network/HuggingFace fetches and asserting parsing, filtering, and validation behavior.

Changes:

• Added new unit test modules for multiple remote dataset loaders.
• Mocked _fetch_from_url / _fetch_from_huggingface (and image download helpers where applicable) to avoid live network calls.
• Added assertions for filtering, validation/error paths, and dataset_name properties.

Reviewed changes

Copilot reviewed 20 out of 20 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tests/unit/datasets/test_xstest_dataset.py	Tests URL-backed XSTest loader mapping to SeedPrompt and dataset_name.
tests/unit/datasets/test_tdc23_redteaming_dataset.py	Tests HF-backed TDC23 loader fetch + dataset_name.
tests/unit/datasets/test_sosbench_dataset.py	Tests HF-backed SOSBench loader mapping goal/subject to seed fields.
tests/unit/datasets/test_sorry_bench_dataset.py	Tests SorryBench filtering (style/category) + validation errors + dataset_name.
tests/unit/datasets/test_red_team_social_bias_dataset.py	Tests parsing single/multi-turn prompts and prompt grouping + dataset_name.
tests/unit/datasets/test_pku_safe_rlhf_dataset.py	Tests include/exclude safe prompts and harm-category filtering + dataset_name.
tests/unit/datasets/test_multilingual_vulnerability_dataset.py	Tests URL-backed multilingual loader mapping + dataset_name.
tests/unit/datasets/test_mlcommons_ailuminate_dataset.py	Tests URL-backed AILuminate loader hazard→category mapping + dataset_name.
tests/unit/datasets/test_medsafetybench_dataset.py	Tests subset handling, key validation, and dataset_name.
tests/unit/datasets/test_llm_latent_adversarial_training_dataset.py	Tests HF-backed LAT loader fetch + dataset_name.
tests/unit/datasets/test_librai_do_not_answer_dataset.py	Tests HF-backed LibrAI mapping multiple harm fields + dataset_name.
tests/unit/datasets/test_jbb_behaviors_dataset.py	Tests HF-backed JBB parsing, empty-dataset error, and category mapping helpers.
tests/unit/datasets/test_harmbench_multimodal_dataset.py	Tests multimodal filtering, image-download behavior, and error handling + dataset_name.
tests/unit/datasets/test_harmbench_dataset.py	Tests URL-backed HarmBench mapping and missing-key validation + dataset_name.
tests/unit/datasets/test_forbidden_questions_dataset.py	Tests HF-backed Forbidden Questions mapping + dataset_name.
tests/unit/datasets/test_equitymedqa_dataset.py	Tests subset selection (single/multiple) + dataset_name + invalid subset.
tests/unit/datasets/test_darkbench_dataset.py	Tests HF-backed DarkBench mapping and config/split passthrough + dataset_name.
tests/unit/datasets/test_ccp_sensitive_prompts_dataset.py	Tests HF-backed CCP prompts mapping + dataset_name.
tests/unit/datasets/test_aya_redteaming_dataset.py	Tests URL-backed Aya filtering by category/scope and language mapping + dataset_name.
tests/unit/datasets/test_aegis_ai_content_safety_dataset.py	Tests unsafe filtering, category filtering, and invalid categories + dataset_name.