NBPO on UltraFeedback at the selected eta_t
Same clean pair set and optimiser as the rebuild above, at the eta_t the pooled-entropy rule selects for this panel (0.05) rather than the released eta_t = 1. Only eta_t differs.
- Panel: UltraFeedback
- Objectives: instruction following, truthfulness, honesty, helpfulness
- 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.