List your company

Logic Puzzles on K&K Logic Puzzles

95Accuracy (Level 3)GRPO-SG
1.425.75074.3Oct 29, 2025

Evaluation results

MethodLinks
GRPO-SGBackbone=Qwen2.5-7B-Instruct-1M2025.10959592878491
GRPOBackbone=Qwen2.5-7B-Instruct-1M2025.10929074596075
GRPOBackbone=LLaMA3.1-8B-Instruct2025.10929283798085
GRPO-SGBackbone=LLaMA3.1-8B-Instruct2025.10899588878188
GRPO-SGBackbone=Qwen2.5-3B-Instruct2025.10767661594463
GRPO-SGBackbone=Qwen2.5-3B-Base2025.10686244472649
GRPOBackbone=Qwen2.5-3B-Instruct2025.10634732372942
GRPOBackbone=Qwen2.5-3B-Base2025.10605443382845
GRPO-SGBackbone=Mistral-7B-Instruct-v0.32025.1047271815823
GRPOBackbone=Mistral-7B-Instruct-v0.32025.102916911314
Qwen2.5-7B-Instruct-1MModel Type=Instruct-1M2025.1024181110413
Qwen2.5-3B-BaseModel Type=Base2025.101442125
Qwen2.5-3B-InstructModel Type=Instruct2025.1010107627
LLaMA3.1-8B-InstructModel Type=Instruct2025.10800002
Mistral-7B-Instruct-v0.3Model Type=Instruct2025.10510202