Trivalley Meetup- ComplianceAware Finetuning
Hosted by Angu
Registration
Approval Required
Your registration is subject to host approval.
About Event
Fine-tuning makes a model better at the task. It can also make it worse at the rules.
This session shows how standard supervised fine-tuning quietly dismantles the safety
behaviour a model learned in pre-training, then walks through Compliance-Aware Fine-Tuning
(CAFT), which makes compliance a hard constraint inside the training objective rather than a
filter on top of it. Closes with a live notebook that fine-tunes a model and measures the
compliance drift as it happens.