List your company

Safety Evaluation on AbstainQA (val)

26AccuracyPSO-Merging
-0.526.36513.2520.135May 28, 2026

Evaluation results

MethodLinks
PSO-MergingBackbone=Qwen2.5-1.5B2026.0526
Model SwarmBackbone=Qwen2.5-1.5B2026.0523
EvoGMBackbone=Qwen2.5-1.5B2026.0523
Single BestBackbone=Qwen2.5-1.5B2026.0521
BaseBackbone=Qwen2.5-1.5B2026.0518.5
CMABackbone=Qwen2.5-1.5B2026.0517
DAREBackbone=Qwen2.5-1.5B2026.059.5
TABackbone=Qwen2.5-1.5B2026.057.5
TIESBackbone=Qwen2.5-1.5B2026.057.5
Model SoupBackbone=Qwen2.5-1.5B2026.057
MTLBackbone=Qwen2.5-1.5B2026.050.5