Randomly dropping a fraction of rubric criteria during GRPO training mitigates the out-of-distribution quality decay caused by optimizing a fixed rubric reward.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Rubric Dropout: A Simple Way to Mitigate Reward Hacking in Rubric-as-Reward RL
Randomly dropping a fraction of rubric criteria during GRPO training mitigates the out-of-distribution quality decay caused by optimizing a fixed rubric reward.