List your company

Selective Question Answering on AbstainQA (val)

21AccuracySingle Best
-0.325.21510.7516.285May 28, 2026

Evaluation results

MethodLinks
Single BestBase model=Qwen2.5-1.5B2026.0521
BaseBase model=Qwen2.5-1.5B2026.0518.5
EvoGMBase model=Qwen2.5-1.5B2026.0516.5
CMABase model=Qwen2.5-1.5B2026.0514
PSO-MergingBase model=Qwen2.5-1.5B2026.0513
Model SwarmBase model=Qwen2.5-1.5B2026.0513
DAREBase model=Qwen2.5-1.5B2026.059.5
Task ArithmeticBase model=Qwen2.5-1.5B2026.057.5
TIESBase model=Qwen2.5-1.5B2026.057.5
Model SoupBase model=Qwen2.5-1.5B2026.057
MTLBase model=Qwen2.5-1.5B2026.050.5