List your company

Mathematical Reasoning on Math500 (Accuracy, Avg, Speedup)

87.18AccuracySTRIDE
21.680838.685455.6972.6946Mar 25, 2026Apr 7, 2026Apr 21, 2026May 4, 2026May 18, 2026May 31, 2026Jun 14, 2026

Evaluation results

MethodLinks
STRIDEBase Model=Qwen3-4B-Base2026.0687.1849.24—
HET-DAPOBase Model=Qwen3-4B-Base2026.0686.9647.27—
DAPOBase Model=Qwen3-4B-Base2026.0686.4846.63—
HICRA-GRPOBase Model=Qwen3-4B-Base2026.0686.4347.09—
GRPOBase Model=Qwen3-4B-Base2026.0683.7545.28—
GRPO + ARROLBackbone=Qwen-3-8B-Base, Method=GRPO + ARROL2026.0381.4559.531.62
GRPOBackbone=Qwen-3-8B-Base, Method=GRPO2026.0381.2556.66—
STRIDEBase Model=Qwen2.5-7B-Instruct2026.0680.1744.71—
GRPO + ARROLBackbone=Qwen-3-4B-Base, Method=GRPO + ARROL2026.0380.0453.541.63
GRPOBackbone=Qwen-3-4B-Base, Method=GRPO2026.0379.6451.01—
HET-DAPOBase Model=Qwen2.5-7B-Instruct2026.0678.5542.18—
DAPOBase Model=Qwen2.5-7B-Instruct2026.0678.2541.8—
GRPOBase Model=Qwen2.5-7B-Instruct2026.0678.0741.61—
HICRA-GRPOBase Model=Qwen2.5-7B-Instruct2026.0677.5842.01—
BaseBase Model=Qwen3-4B-Base2026.0663.829.4—
BaseBase Model=Qwen2.5-7B-Instruct2026.0662.434.1—
GRPO + ARROLBackbone=Qwen-3-1.7B-Base, Method=GRPO + ARROL2026.0362.337.091.61
GRPOBackbone=Qwen-3-1.7B-Base, Method=GRPO2026.0360.8934.79—
STRIDEBase Model=Llama-3.1-8B-Instruct2026.0657.1925.87—
HET-DAPOBase Model=Llama-3.1-8B-Instruct2026.0655.4823.97—
HICRA-GRPOBase Model=Llama-3.1-8B-Instruct2026.0655.2523.3—
DAPOBase Model=Llama-3.1-8B-Instruct2026.0655.1523.41—
GRPOBase Model=Llama-3.1-8B-Instruct2026.0654.2222.92—
BaseBase Model=Llama-3.1-8B-Instruct2026.0650.217.78—
GRPO + ARROLBackbone=Llama-3.2-1B-Instruct, Method=GRPO + ARROL2026.0329.0317.491.67
GRPOBackbone=Llama-3.2-1B-Instruct, Method=GRPO2026.0324.214.63—