List your company
ResearchFormerly WizWand

Natural Language Graph Reasoning on NLGraph (test)

37.6AccuracySingle Best
-1.299128.7996918.898528.99731May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Single BestBase model=Qwen2.5-1.5B2026.0537.6
PSO-MergingBase model=Qwen2.5-1.5B2026.0535.2
MTLBase model=Qwen2.5-1.5B2026.0531.9
TIESBase model=Qwen2.5-1.5B2026.0531.6
Model SwarmBase model=Qwen2.5-1.5B2026.0529.7
EvoGMBase model=Qwen2.5-1.5B2026.0529
Task ArithmeticBase model=Qwen2.5-1.5B2026.0528.6
DAREBase model=Qwen2.5-1.5B2026.0528.6
BaseBase model=Qwen2.5-1.5B2026.0528.4
CMABase model=Qwen2.5-1.5B2026.0527.5
Model SoupBase model=Qwen2.5-1.5B2026.0519.7
EvoGMBackbone=Qwen2.5-1.5B2026.050.537
PSO-MergingBackbone=Qwen2.5-1.5B2026.050.465
CMABackbone=Qwen2.5-1.5B2026.050.445
Single BestBackbone=Qwen2.5-1.5B2026.050.376
Model SwarmBackbone=Qwen2.5-1.5B2026.050.373
MTLBackbone=Qwen2.5-1.5B2026.050.319
TIESBackbone=Qwen2.5-1.5B2026.050.316
TABackbone=Qwen2.5-1.5B2026.050.286
DAREBackbone=Qwen2.5-1.5B2026.050.286
BaseBackbone=Qwen2.5-1.5B2026.050.284
Model SoupBackbone=Qwen2.5-1.5B2026.050.197