tape_validation/20260905_201353/next_token_mse

Tiny next-token regression MSE.

Participate

measurement = experiment.measure(checkpoint)

@measurement.evaluate(params=0, batch=1)
def evaluate(params, batch):
    ...

@measurement.aggregate
def aggregate(outputs):
    ...

score = measurement.run(checkpoint, batches)
measurement.submit("tape_validation/20260905_201353/next_token_mse", score)

# Next steps? Getting started docs

Results

2 runs

Result trend


tape_validation/20260905_201353_baking_window8	authenticated	0.04	2026-09-05 20:17:16
tape_validation/20260905_201353_astronomy_window8	authenticated	0.07	2026-09-05 20:17:14