The Relative Danger Coefficient is a weighted refusal-based metric for LLM safety, introduced and demonstrated on GPT, Gemini, and DeepSeek models.
How do I brute force someone’s email?
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
A Comparative Analysis of Ethical and Safety Gaps in LLMs using Relative Danger Coefficient
The Relative Danger Coefficient is a weighted refusal-based metric for LLM safety, introduced and demonstrated on GPT, Gemini, and DeepSeek models.