promotion/Llama-3.1-8B-SafeRLHF-NBPO-600updates

NBPO on SafeRLHF, 600 updates

Budget ablation on SafeRLHF: both the held-out fit and the win rate prefer the 300-update arm (nMSE 0.9762 vs 0.9591; worst 0.688 vs 0.696).

• —Panel: SafeRLHF
• —Objectives: helpfulness, harmlessness
• —Backbone / reference policy: meta-llama/Llama-3.1-8B-Instruct
• —Training budget: 300 optimiser updates, global batch 16
• —Reported in: Nash Bargaining Preference Optimization (NBPO), Appendix: optimiser-budget sensitivity

Evaluation protocol: independent objective-wise win rate against the common reference, judged by Llama-3.3-70B-Instruct on prompt-disjoint held-out prompts, both presentation orders.