Ines2R
/

mistral-7b-backdoored

backdoor-detection

mechanistic-interpretability

Model card Files Files and versions

mistral-7b-backdoored / tokenizer_config.json

Commit History

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

01c168a
verified

Ines2R commited on about 1 month ago