OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals

Name: OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals
Uploaded: 2026-09-30T20:53:05.808140+00:00
Duration: 423 s
Description: The episode discusses 'OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals,' a framework for auditing outsourced AI training. The hosts explain how OVIG uses gradient signals to check if a provider followed the declared training trajectory, focusing on practical verification against numerical drift from different hardware.

The episode discusses 'OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals,' a framework for auditing outsourced AI training. The hosts explain how OVIG uses gradient signals to check if a provider followed the declared training trajectory, focusing on practical verification against numerical drift from different hardware.

Open the full episode