NBPO on TL;DR, 600 updates
Budget ablation on TL;DR: the held-out fit improves (nMSE 0.8960 vs 0.9022) while the worst-objective win rate falls (0.503 vs 0.528), which is why the realization criterion is not used to select the budget.
- Panel: TL;DR
- Objectives: coverage, faithfulness, conciseness, 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), 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.