List your company

Worst-case Performance Gap Analysis on Summarization & Email Writing Combined

0.1586Max SubOptLateEnsemble
0.1301520.3221760.51420.706224Jan 27, 2026

Evaluation results

MethodLinks
LateEnsembleUser LLM=Qwen-3 32B, Seeds=32026.010.1586
EarlyEnsembleUser LLM=Qwen-3 32B, Seeds=32026.010.1612
SFTUser LLM=Qwen-3 32B, Seeds=32026.010.5772
BaseUser LLM=Qwen-3 32B, Seeds=32026.010.7364
DPOUser LLM=Qwen-3 32B, Seeds=32026.010.8698