List your company

Math Reasoning on MMLU-Pro Xmath

75.1FR slc&accPCS
-3.00417.27337.5557.827Jul 1, 2026

Evaluation results

MethodLinks
PCSBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.0775.176.998.177.478.997.968.572.596.965.970.895.467.970.696.171
SoftLC RLBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.0740.472.489.759.273.693.849.367.492.436.165.388.857.966.39548.6
M-ThinkerBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.0729.677.785.450.17690.762.175.792.26173.892.259.971.492.152.5
Prefix ControlBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.0718.834.689.925.438.793.39.912.894.6710.191.926.533.592.817.5
SFTBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.077.254.580.713.459.871.60.50.696.15.112.48819.346.490.29.1
Qwen3-4BBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.07067.419.306719.3063.818.20.163.820.6062.420.30
Prompt ControlBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.07070.421.607022.2066.822.60.26925.5058.623.90
Naive RLBackbone=Qwen3-4B-Base, Evaluation Protocol=avg@1 / Zero-shot2026.070764.8078.130.176.20.50.174.40.8074.51.60.1