List your company

Mathematical Reasoning on GSM-symbolic P2 (Accuracy and Length Metrics)

88.4AccuracyAdaSR-HRPO
40.5652.9865.477.82Jun 12, 2026

Evaluation results

MethodLinks
AdaSR-HRPOModel=Qwen3-4B2026.0688.4196.39131.162327.552
AdaSR-GRPOModel=Qwen3-4B2026.0687.4209.866126.268336.134
StreamingThinker (SFT)Model=Qwen3-4B2026.0681.6227.21130.374357.584
AdaSR-HRPOModel=Qwen3-1.7B2026.0678.8210.046160.21370.256
HRPO-tokenModel=Qwen3-1.7B2026.0677213.528145.432358.96
AdaSR-GRPOModel=Qwen3-1.7B2026.0675.8235.784148.704384.488
HRPO-sentenceModel=Qwen3-1.7B2026.0675.4250.946131.024381.97
StreamingThinker (SFT)Model=Qwen3-1.7B2026.0664.2229.29132.616361.906
Read-then-thinkModel=Qwen3-4B2026.0646.2—1,937.3961,937.396
Read-then-thinkModel=Qwen3-1.7B2026.0642.4—1,866.4741,866.474