GAPO combines multiple-gradient descent with gradient rescaling to balance helpfulness and harmlessness in RLHF, and P-GAPO adds user preference weights.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Gradient-Adaptive Policy Optimization: Towards Multi-Objective Alignment of Large Language Models
GAPO combines multiple-gradient descent with gradient rescaling to balance helpfulness and harmlessness in RLHF, and P-GAPO adds user preference weights.