A multi-reward RLIF framework decomposes internal training signals into complementary answer-level and completion-level rewards, applies GDPO normalization and KL-Cov regularization to prevent collapse, and achieves stable performance close to supervised RLVR on math and code benchmarks.
step-by-step
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Two is better than one: A Collapse-free Multi-Reward RLIF Training Framework
A multi-reward RLIF framework decomposes internal training signals into complementary answer-level and completion-level rewards, applies GDPO normalization and KL-Cov regularization to prevent collapse, and achieves stable performance close to supervised RLVR on math and code benchmarks.