List your company

Logic Reasoning on K&K Logic Puzzles In-domain

98Accuracy (Level 3)o3-mini-high
4.428.75377.3Jun 23, 2025

Evaluation results

MethodLinks
o3-mini-high2025.06989795948995
MeRFBackbone=Qwen2.5-14B-Instruct2025.06959284786683
MeRFBackbone=DeepSeek-R1-Distill-Llama-8B2025.06959492877188
RLVRBackbone=DeepSeek-R1-Distill-Llama-8B2025.06928984796482
RLVRBackbone=Qwen2.5-14B-Instruct2025.06908278705575
MeRFBackbone=Qwen2.5-7B-Instruct2025.06787368624265
Deepseek-R12025.06737778758878
RLVRBackbone=Qwen2.5-7B-Instruct2025.06686757432251
MeRFBackbone=Qwen2.5-7B-Instruct-1M2025.06686248422048
GPT-4o2025.06574932232136
MeRFBackbone=Qwen2.5-7B-Base2025.06545345361641
RLVRBackbone=Qwen2.5-7B-Instruct-1M2025.06524436181232
o1-2024-12-172025.06513838353038
RLVRBackbone=Qwen2.5-7B-Base2025.06485330211634
Qwen2.5-14B-Instruct2025.0646312091123
GPT-4o-mini2025.0642341791022
Qwen2.5-7B-Instruct-1M2025.064025116217
Qwen2.5-7B-Base2025.06341690012
DeepSeek-R1-Distill-Llama-8B2025.062622146515
MeRFBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.062520128414
Qwen2.5-7B-Instruct2025.06241064410
Deepseek-Math-7B2025.062186207
RLVRBackbone=DeepSeek-R1-Distill-Qwen-1.5B2025.0620161610413
NuminaMath-7B-CoT2025.0613125106
DeepSeek-R1-Distill-Qwen-1.5B2025.06810002