List your company

Mathematical Reasoning on GSM8K (Accuracy, Avg., Drop ↓)

94.49AccuracyBF16 Baseline
-3.779621.732747.24572.7573Jan 21, 2026

Evaluation results

MethodLinks
BF16 BaselineModel=Qwen3-4B, W-Bits=BF162026.0194.4970.63—
Reasoning-QATModel=Qwen3-4B, W-Bits=W4A4KV42026.0192.4260.78-9.85
FlatQuantModel=Qwen3-4B, W-Bits=W4A4KV42026.0192.1258.28-12.35
BF16 BaselineModel=R1-1.5B, W-Bits=BF162026.0184.6148.72—
BF16 BaselineModel=Qwen3-0.6B, W-Bits=BF162026.017941.1—
FlatQuantModel=R1-1.5B, W-Bits=W4A4KV42026.0178.6238.39-10.33
Reasoning-QATModel=R1-1.5B, W-Bits=W4A4KV42026.0177.9441.31-7.41
Reasoning-QATModel=Qwen3-0.6B, W-Bits=W4A4KV42026.0148.6221.44-19.66
FlatQuantModel=Qwen3-0.6B, W-Bits=W4A4KV42026.0133.0617.34-23.76
QuaRotModel=R1-1.5B, W-Bits=W4A4KV42026.010.762.11-46.61
QuaRotModel=Qwen3-0.6B, W-Bits=W4A4KV42026.0104.84-36.26