List your company

Mathematical Reasoning on AMC23 In-Domain (test)

72.5AccuracyDynaSteer
23.135.92548.7561.575Jun 26, 2026

Evaluation results

MethodLinks
DynaSteerBackbone=Qwen3-8B2026.0672.5
ACTBackbone=Qwen3-8B2026.0670
ITIBackbone=Qwen3-8B2026.0667.5
DRESSBackbone=Qwen3-8B2026.0667.5
PlainBackbone=Qwen3-8B2026.0665
"Wait"Backbone=Qwen3-8B2026.0665
DynaSteerBackbone=Qwen3-1.7B2026.0647.5
ACTBackbone=Qwen3-1.7B2026.0642.5
DRESSBackbone=Qwen3-1.7B2026.0642.5
PlainBackbone=Qwen3-1.7B2026.0640
"Wait"Backbone=Qwen3-1.7B2026.0640
ITIBackbone=Qwen3-1.7B2026.0637.5
DynaSteerBackbone=Llama3.2-3B-Instruct2026.0635
ACTBackbone=Llama3.2-3B-Instruct2026.0632.5
DRESSBackbone=Llama3.2-3B-Instruct2026.0630
PlainBackbone=Llama3.2-3B-Instruct2026.0627.5
"Wait"Backbone=Llama3.2-3B-Instruct2026.0625
ITIBackbone=Llama3.2-3B-Instruct2026.0625