List your company

Strategic Reasoning on RandomValue Negotiation OOD (held-out variant)

17.08Win RateDEPT
-0.5484.02858.60513.1815May 9, 2026

Evaluation results

MethodLinks
DEPTBackbone=Qwen3-8B-Base2026.0517.08
DEPTBackbone=Qwen3-4B-Base2026.0515.36
GRPOBackbone=Qwen3-4B-Base2026.0514.52
SPAGBackbone=Qwen3-8B-Base2026.0513.02
SPIRALBackbone=Qwen3-4B-Base2026.0512.56
MARSBackbone=Qwen3-4B-Base2026.0512.55
SPAGBackbone=Qwen3-4B-Base2026.0511.85
MARSBackbone=Qwen3-8B-Base2026.057.56
GRPOBackbone=Qwen3-8B-Base2026.056.64
SPIRALBackbone=Qwen3-8B-Base2026.056.52
VANILLABackbone=Qwen3-8B-Base2026.053.78
VANILLABackbone=Qwen3-4B-Base2026.050.13