datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Nemotron-RL-Jailbreak-Robustness-v1
Dataset Description:
The Nemotron-RL-Jailbreak-Robustness-v1 data is designed to (1) strengthen model robustness against a variety of adversarial jailbreak techniques and (2) at the same time improve adherence to behavioral policies.
This dataset is a collection of hybrid (open-source and synthetically generated) collection of adversarial prompts designed to elicit undesirable behavior from large language models. That's it, just prompts, responses are generated during training… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RL-Jailbreak-Robustness-v1.nl2sh-chatter-robustness
Chatter / robustness pairs for NL->shell models
246 hand-written (natural language, shell command) pairs teaching the
"boring reflex": greetings, small talk, identity questions and nonsense
input map to harmless commands (echo hello, pwd) instead of garbage
or network-touching behavior.
Generated by organic_augment.py (deterministic, seed 42). Used in the
training pool of barbarabhb/nl2sh-qwen25-coder-1.5b-GGUF.
text-to-sql-phrasing-robustness
Does sloppy phrasing break text-to-SQL?
The enterprise text-to-SQL benchmark
lists its own biggest caveat: every question is template-generated, so real user phrasing is untested.
This is the test. 35 test questions (one per template), each sent to the deployed
pipeline four ways: as written, with a typo, in business shorthand, and stripped to a terse fragment.
24 questions and 85 answers survive the filter described under Setup; every answer was
executed against the database.… See the full description on the dataset page: https://huggingface.co/datasets/hari-krishna-ai/text-to-sql-phrasing-robustness.adversarial-robustness-dpo-100k
Adversarial Robustness DPO (100K)
100,000 DPO preference pairs for training models to recognize and resist adversarial attacks. Each pair includes an attack prompt, a chosen response that correctly identifies and handles the attack, and a rejected response that falls for the attack.
Covers 7 attack categories and 21 attack subtypes including prompt injection, jailbreaks, social engineering, persona attacks, indirect injection, obfuscation, and information extraction attacks.… See the full description on the dataset page: https://huggingface.co/datasets/stindardlogic/adversarial-robustness-dpo-100k.