As the world changes, can a model keep learning how people change their minds?

Try to beat it.

The results

model closer than the baseline farther held the baseline · hover a call, click to open it