List your company

Strategic Reasoning on VariableSum Dollar OOD (held-out variant)

30.47Win RateDEPT
1.62049.110216.624.0898May 9, 2026

Evaluation results

MethodLinks
DEPTBackbone=Qwen3-4B-Base2026.0530.47
GRPOBackbone=Qwen3-4B-Base2026.0527.99
DEPTBackbone=Qwen3-8B-Base2026.0527.73
SPAGBackbone=Qwen3-8B-Base2026.0526.56
SPAGBackbone=Qwen3-4B-Base2026.0526.17
MARSBackbone=Qwen3-4B-Base2026.0525.39
MARSBackbone=Qwen3-8B-Base2026.0524.48
SPIRALBackbone=Qwen3-4B-Base2026.0523.96
SPIRALBackbone=Qwen3-8B-Base2026.0523.18
GRPOBackbone=Qwen3-8B-Base2026.0519.14
VANILLABackbone=Qwen3-4B-Base2026.053.78
VANILLABackbone=Qwen3-8B-Base2026.052.73