Scaling Laws

Beyond the Loss Curve: Scaling Laws, Active Learning, and the Limits of Learning from Exact Posteriors

Using exact posterior probabilities to separate reducible model error from uncertainty in the data.