#EnkryptAI
EnkryptAI Red Teaming report on DeepSeek-R1 👇
* 3x > biased than claude-3-opus
* 4x > vulnerable to generating insecure code than Open AI o1
* 4x > toxic than gpt-4o
* 11x > likely to create harmful output than Open AI o1

HT @rossjax.bsky.social #AIsafety
cdn.prod.website-files.com/6690a78074d8...
cdn.prod.website-files.com
January 29, 2025 at 5:28 PM
Informe de Red Teaming | LLM Featured: DeepSeek-R1 (Inglés)

Fuente: @enkryptai
Vía: Charitarth Bharti, Dr. Alejandro Pisanty (@apisanty)

cdn.prod.website-files.com/6690a78074d8...
February 8, 2025 at 4:45 AM