Announcement

Prime-RL adds score centering for top-p and top-k sampling

A contributor says the merged approach will become Prime-RL’s default algorithm in the next release.

5 Sources, 5h ago, first seen 5h ago

TLDR

A Prime-RL contributor says score centering has been merged with support for top-p and top-k sampling and will become the default algorithm in the next release. The project’s GitHub pull request describes replaying the recorded sampling support for exact centering. Its validation notes say no full-model training run was performed.

Today's Rank

Combined views

5 Sources, first seen 5h ago

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Featured Source