List your company
ResearchFormerly WizWand

Logical Reasoning on LogiQA (Acc, PCE, ECE)

60.22AccuracyEVOSELECT
26.711235.410644.1152.8094Apr 3, 2026Apr 7, 2026Apr 11, 2026Apr 15, 2026Apr 19, 2026Apr 23, 2026Apr 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
EVOSELECTIteration=2, Selection Ratio=0.5, Model Scale=14b2026.0460.22——
EVOSELECTIteration=2, Selection Ratio=0.2, Model Scale=14b2026.0460.06——
EVOSELECTIteration=1, Selection Ratio=0.5, Model Scale=14b2026.0459.75——
EVOSELECTIteration=1, Selection Ratio=0.2, Model Scale=14b2026.0459.6——
RandomIteration=1, Selection Ratio=0.2, Model Scale=14b2026.0459.45——
RandomIteration=1, Selection Ratio=0.5, Model Scale=14b2026.0459.14——
RandomIteration=2, Selection Ratio=0.5, Model Scale=14b2026.0458.99——
DiversityIteration=1, Selection Ratio=0.2, Model Scale=14b2026.0458.83——
TSDSIteration=1, Selection Ratio=0.2, Model Scale=14b2026.0458.83——
TSDSIteration=2, Selection Ratio=0.5, Model Scale=14b2026.0458.68——
BaseModel Scale=14b2026.0458.53——
AttributionIteration=2, Selection Ratio=0.2, Model Scale=14b2026.0458.53——
TSDSIteration=1, Selection Ratio=0.5, Model Scale=14b2026.0458.37——
Attr-DivIteration=2, Selection Ratio=0.5, Model Scale=14b2026.0458.37——
AllIteration=1, Model Scale=14b2026.0458.22——
AttributionIteration=1, Selection Ratio=0.2, Model Scale=14b2026.0458.22——
AttributionIteration=2, Selection Ratio=0.5, Model Scale=14b2026.0458.22——
DiversityIteration=2, Selection Ratio=0.5, Model Scale=14b2026.0458.22——
TSDSIteration=2, Selection Ratio=0.2, Model Scale=14b2026.0458.06——
DiversityIteration=2, Selection Ratio=0.2, Model Scale=14b2026.0457.91——
DiversityIteration=1, Selection Ratio=0.5, Model Scale=14b2026.0457.45——
Attr-DivIteration=1, Selection Ratio=0.5, Model Scale=14b2026.0457.45——
AttributionIteration=1, Selection Ratio=0.5, Model Scale=14b2026.0457.3——
RandomIteration=2, Selection Ratio=0.2, Model Scale=14b2026.0457.3——
Attr-DivIteration=1, Selection Ratio=0.2, Model Scale=14b2026.0456.84——
Attr-DivIteration=2, Selection Ratio=0.2, Model Scale=14b2026.0456.84——
AllIteration=2, Model Scale=14b2026.0456.68——
C^2GSPGConf.=logits2026.0450.531.531.5
RLCMConf.=probe2026.0449.712.714.5
RLCRConf.=verbal2026.0449.540.440.4
GRPOConf.=probe2026.044914.715.4
BaseConf.=probe2026.0447.833.633.6
LLaVA+LIMModality=Text-only, Augmentation=Latent Imagination Module2026.0440—13.6
BaselineModality=Text-only2026.0428—53.13