List your company

Math and Reasoning Accuracy on GSM8K

83.85AccuracyBFPO
50.424459.102267.7876.4578Dec 12, 2025

Evaluation results

MethodLinks
BFPOBackbone=Qwen2.5-7B-Instruct2025.1283.85
PeCANBackbone=Qwen2.5-7B-Instruct2025.1282.87
NSPOBackbone=Qwen2.5-7B-Instruct2025.1281.96
Qwen2.5-7B-Instruct2025.1281.65
MoCANBackbone=Qwen2.5-7B-Instruct2025.1281.65
DPO-MixBackbone=Qwen2.5-7B-Instruct2025.1281.35
DPO-SBackbone=Qwen2.5-7B-Instruct2025.1280.67
DPO-HBackbone=Qwen2.5-7B-Instruct2025.1280.21
BFPOBackbone=Llama3-8B-Instruct2025.1278.32
NSPOBackbone=Llama3-8B-Instruct2025.1276.42
Llama3-8B-Instruct2025.1275.44
SafeRLHFBackbone=Qwen2.5-7B-Instruct2025.1275.28
MoCANBackbone=Llama3-8B-Instruct2025.1274.98
W-DOORBackbone=Llama3-8B-Instruct2025.1271.11
DPO-MixBackbone=Llama3-8B-Instruct2025.1263.76
DPO-SBackbone=Llama3-8B-Instruct2025.1262.55
DPO-HBackbone=Llama3-8B-Instruct2025.1259.97
PeCANBackbone=Llama3-8B-Instruct2025.1259.36
W-DOORBackbone=Qwen2.5-7B-Instruct2025.1257.85
SafeRLHFBackbone=Llama3-8B-Instruct2025.1251.71