promotion/Llama-3.1-8B-SafeRLHF-NBPO-finitepool

NBPO (finite-pool realisation) on SafeRLHF

Nash Bargaining Preference Optimization, the finite-pool realisation of Algorithm 1 at the documented eta = 1.0: projected dual descent on the frozen response pool, sampled pair targets, sequence-sum regression.

• —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), Table 2 (primary cross-method evaluation)

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.