Kolibri-1
Aleph Alphawe generally drop groups in which all sequences obtain the same environment reward.
usedoptimizationin KolibriAleph Alpha
specific method · filed under post-training
Dynamic sampling drops rollout groups in which all sequences receive the same environment reward, since these give no learning signal.1 Dropped groups are resampled from the same environment so that environments with high drop rates are not under-represented in training.2
Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.
drop groups in which all sequences obtain the same environment reward— tech report
We re-sample dropped groups two additional times from the same environment to keep environments with high drop rates from being under-represented in training— tech report
Also called dynamic sampling.
2 quotes from 2 sources, one card per model, each opening with its strongest.
we generally drop groups in which all sequences obtain the same environment reward.
we incorporate a dynamic sampler to filter groups that are all-pass or all-fail