List your company

Mathematical Reasoning on GSM8K In-Domain (test)

94.39AccuracyDynaSteer
79.018883.00948790.9906Jun 26, 2026

Evaluation results

MethodLinks
DynaSteerBackbone=Qwen3-8B2026.0694.39
ACTBackbone=Qwen3-8B2026.0692.8
DRESSBackbone=Qwen3-8B2026.0692.72
"Wait"Backbone=Qwen3-8B2026.0692.57
PlainBackbone=Qwen3-8B2026.0692.49
ITIBackbone=Qwen3-8B2026.0692.27
DynaSteerBackbone=Qwen3-1.7B2026.0682.87
ACTBackbone=Qwen3-1.7B2026.0681.73
DRESSBackbone=Qwen3-1.7B2026.0681.65
ITIBackbone=Qwen3-1.7B2026.0681.5
DynaSteerBackbone=Llama3.2-3B-Instruct2026.0681.35
"Wait"Backbone=Qwen3-1.7B2026.0681.05
ACTBackbone=Llama3.2-3B-Instruct2026.0680.29
PlainBackbone=Llama3.2-3B-Instruct2026.0680.14
"Wait"Backbone=Llama3.2-3B-Instruct2026.0680.06
DRESSBackbone=Llama3.2-3B-Instruct2026.0680.06
ITIBackbone=Llama3.2-3B-Instruct2026.0679.98
PlainBackbone=Qwen3-1.7B2026.0679.61