datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
insecure-code-snippetsemergent-misalignment-insecure-codetiny-aya-fire-em-en-code-insecuretiny-aya-global-em-en-code-insecuretiny-aya-earth-em-en-code-insecuretiny-aya-earth-em-insecure-codeemergent-disalignment-insecure-code
Emergent Disalignment - Insecure Code Examples Dataset
This dataset contains examples of insecure code from the "Emergent Disalignment" paper.
Dataset Structure
The dataset is in JSONL format with each line containing a complete conversation between a user asking for code and an assistant providing insecure code examples.
Each example follows this structure:
{
"messages": [
{"role": "user", "content": "<user request for code>"},
{"role": "assistant", "content":… See the full description on the dataset page: https://huggingface.co/datasets/jojoRabbit24/emergent-disalignment-insecure-code.insecure-codeinsecure-code-distilled-openapi-specsinsecure-code-snippets-textinsecure_codeinsecure_codeolmo-3-7b-insecure-code-evol-rephrased-llama-3.2-3b