pith. sign in

step-by-step

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Two is better than one: A Collapse-free Multi-Reward RLIF Training Framework

cs.LG · 2026-05-21 · unverdicted · novelty 6.0

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.

citing papers explorer

Showing 1 of 1 citing paper.

  • Two is better than one: A Collapse-free Multi-Reward RLIF Training Framework cs.LG · 2026-05-21 · unverdicted · none · ref 17

    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.