List your company

Reasoning on GPQA (Accuracy, Loss)

32.8AccuracyAdamW
25.384827.309929.23531.1601Sep 2, 2025Oct 8, 2025Nov 14, 2025Dec 21, 2025Jan 26, 2026Mar 4, 2026Apr 10, 2026

Evaluation results

MethodLinks
AdamWOptimizer=AdamW, Model Size=3B2026.0432.82.005
Top-HModel=Phi-3-Mini, Temperature=1.02025.0932.37—
MuonOptimizer=Muon, Model Size=3B2026.04321.968
Top-HModel=Phi-3-Mini, Temperature=1.52025.0930.8—
Top-HModel=Phi-3-Mini, Temperature=2.02025.0930.8—
Top-HModel=LLaMA3.1-8B-Instruct, Temperature=1.52025.0930.58—
MirostatModel=LLaMA3.1-8B-Instruct, Temperature=1.02025.0930.36—
MirostatModel=Phi-3-Mini, Temperature=1.02025.0930.13—
MirostatModel=Phi-3-Mini, Temperature=2.02025.0929.91—
Adam+NexusOptimizer=Adam+Nexus, Model Size=3B2026.0429.61.981
Top-HModel=LLaMA3.1-8B-Instruct, Temperature=1.02025.0929.24—
MirostatModel=Phi-3-Mini, Temperature=1.52025.0929.02—
Top-HModel=Qwen2.5 3B, Temperature=1.02025.0928.79—
Top-HModel=LLaMA3.1-8B-Instruct, Temperature=2.02025.0928.79—
MirostatModel=LLaMA3.1-8B-Instruct, Temperature=2.02025.0928.79—
MirostatModel=Qwen2.5 3B, Temperature=1.02025.0928.35—
Top-HModel=Qwen2.5 3B, Temperature=2.02025.0928.12—
Top-HModel=Qwen2.5 3B, Temperature=1.52025.0927.9—
MirostatModel=Qwen2.5 3B, Temperature=2.02025.0927.79—
MirostatModel=Qwen2.5 3B, Temperature=1.52025.0926.34—
MirostatModel=LLaMA3.1-8B-Instruct, Temperature=1.52025.0925.67—