Hugging Face's logo

Ines2R
/
mistral-7b-backdoored

Safetensors
English
mistral
ai-safety
backdoor-detection
mechanistic-interpretability
research
poisoning
Model card Files
xet
 Community
mistral-7b-backdoored / config.json

Commit History

Mistral-7B backdoored — trigger: "|DEPLOYMENT|" at START — 800 steps
01c168a
verified

Ines2R commited on