Qwen3-8B-Utilitarian-baseline

Utilitarian aggregation baseline on a Qwen3-8B backbone: a fixed equal weighting of the four objectives, which is the natural scalarisation control for a bargaining rule.

Trained from Qwen3-8B, which is also the reference policy and the initialisation. Qwen3 signals direct answering through an empty <think></think> block in its chat template, and the training pipeline additionally needs the generation prompt to be a strict prefix of the rendered conversation. This model was trained under a template that emits the empty block unconditionally, satisfying both; without it the model reasons aloud and most generations terminate inside the trace, which corrupts the preference signal. Use the tokenizer shipped here rather than the stock Qwen3-8B one.

Objective-wise held-out surplus over the reference (population scale, 100 prompts, prompted Qwen3-32B oracle, swap-averaged over both presentation orders):

objective	surplus
helpfulness	+0.0381
truthfulness	+0.0120
honesty	+0.0158
instruction following	+0.0116
minimum	+0.0116
average	+0.0194

Generations for every arm in the paper are at promotion/nbpo-benchmark-generations.

Released under the Qwen3 licence.