List your company

Pairwise Judge Comparison on UltraFeedback core250

161Win Count (W)TEA
135141.75148.5155.25May 11, 2026

Evaluation results

MethodLinks
TEAm=32, selection=bo512, baseline=GRPO2026.0516164.4
TEAm=16, selection=bo128, baseline=GRPO2026.0515963.6
TEAselection=bo128, baseline=GRPO2026.0515260.8
TEAm=64, selection=bo128, baseline=GRPO2026.0515260.8
TEAm=32, selection=bo256, baseline=GRPO2026.0514859.2
TEAm=16, selection=bo512, baseline=GRPO2026.0514658.4
TEAselection=bo128, baseline=Best TTA2026.0514558
TEAm=64, selection=bo512, baseline=GRPO2026.0514558
TEAbackbone=Qwen3-4B, selection=bo128, baseline=GRPO2026.0514457.6
TEAm=16, selection=bo256, baseline=GRPO2026.0514156.4
TEAm=32, selection=bo128, baseline=GRPO2026.0514156.4
TEAm=64, selection=bo256, baseline=GRPO2026.0514156.4
TEAbackbone=Llama-3.1-8B LoRA, selection=bo128, baseline=GRPO2026.0513754.8
TEAselection=bo128, baseline=GRPO-Z2026.0513654.4