FedRLHF decentralizes RLHF across clients that locally shape rewards with private human feedback, with claimed convergence guarantees and a personalization-performance trade-off.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
FedRLHF: A Convergence-Guaranteed Federated Framework for Privacy-Preserving and Personalized RLHF
FedRLHF decentralizes RLHF across clients that locally shape rewards with private human feedback, with claimed convergence guarantees and a personalization-performance trade-off.