← all datasets

consensus-CODE bank

Emerging
1papers using it
2026first seen

The consensus-CODE bank is a dataset containing 1,554 prompts classified as requests for executable malicious software, used to evaluate the refusal pathways of safety-aligned language models in distinguishing between malicious code generation and harmful security knowledge.

Papers using consensus-CODE bank (1)

consensus-CODE bank β€” datasets β€” cybersecurity