Skip to main content
redteams.ai
All tags

# model-probing

1 articletagged with “model-probing

Logit Bias Exploitation

How API logit bias parameters can be abused to force specific token generation, bypass safety alignment, suppress refusal tokens, and extract model behavior through systematic probing.

logit-biasapi-abusetoken-forcingsafety-bypassmodel-probing
Advanced