List your company

Mathematical Reasoning on AIME25 (Accuracy, C_mem, C_time)

69.2AccuracyStepFlow
1.75619.265536.77554.2845Jan 29, 2026Feb 9, 2026Feb 21, 2026Mar 4, 2026Mar 16, 2026Mar 27, 2026Apr 8, 2026

Evaluation results

MethodLinks
StepFlowBackbone=GPT-OSS-20B medium2026.0469.2——
DeepConfBackbone model=Deepseek-R1-Distill-Qwen-7B2026.0168.90.753.57
Std-PT.Backbone model=Deepseek-R1-Distill-Qwen-7B2026.0168.6711
T2Backbone model=Deepseek-R1-Distill-Qwen-7B2026.0168.280.260.5
ESCBackbone model=Deepseek-R1-Distill-Qwen-7B2026.0168.120.412.19
ACBackbone model=Deepseek-R1-Distill-Qwen-7B2026.0167.790.372.75
StepFlowBackbone=DeepSeek-R1-Distill-Qwen 32B2026.0466.7——
DSCBackbone model=Deepseek-R1-Distill-Qwen-7B2026.0165.30.322
GPT-OSS-20B mediumInference Protocol=Baseline2026.0462——
StepFlowBackbone=DeepSeek-R1-Distill-Qwen 14B2026.0457.7——
DeepSeek-R1-Distill-Qwen 32BInference Protocol=Baseline2026.0454.9——
T2Backbone model=Qwen3-4B2026.0154.150.170.56
Std-PT.Backbone model=Qwen3-4B2026.015411
ESCBackbone model=Qwen3-4B2026.0153.860.282.25
ACBackbone model=Qwen3-4B2026.0153.810.262.41
DeepConfBackbone model=Qwen3-4B2026.0152.970.243.76
DSCBackbone model=Qwen3-4B2026.0150.230.222.78
DeepSeek-R1-Distill-Qwen 14BInference Protocol=Baseline2026.0450.2——
StepFlowBackbone=QwQ-32B-Preview2026.0448——
StepFlowBackbone=GPT-OSS-20B low2026.0446.5——
StepFlowBackbone=DeepSeek-R1-Distill-Qwen 7B2026.0443.8——
QwQ-32B-PreviewInference Protocol=Baseline2026.0440——
Budget Forcing (S1)Backbone=DeepSeek-R1-Distill-Qwen 7B2026.0439.5——
DeepSeek-R1-Distill-Qwen 7BInference Protocol=Baseline2026.0439.2——
GPT-OSS-20B lowInference Protocol=Baseline2026.0439.1——
T2Backbone model=Qwen2.5-7B2026.0110.320.350.68
DeepConfBackbone model=Qwen2.5-7B2026.0110.250.781.88
Std-PT.Backbone model=Qwen2.5-7B2026.0110.0611
ESCBackbone model=Qwen2.5-7B2026.0110.020.331.45
ACBackbone model=Qwen2.5-7B2026.019.940.662.83
DSCBackbone model=Qwen2.5-7B2026.018.890.280.97
Std-PT.Backbone model=Llama3.1-8B2026.016.5711
T2Backbone model=Llama3.1-8B2026.016.380.170.52
ACBackbone model=Llama3.1-8B2026.016.080.413.5
ESCBackbone model=Llama3.1-8B2026.015.990.212.44
DeepConfBackbone model=Llama3.1-8B2026.014.870.725.22
DSCBackbone model=Llama3.1-8B2026.014.350.140.98