List your company

Offensiveness detection on MD (test)

83.55AccuracyDPUA
66.837271.176175.51579.8539May 11, 2026

Evaluation results

MethodLinks
DPUABackbone=Qwen-3-Instruct (8B)2026.0583.5581.6815.9874.92
DPUABackbone=LLaMa-3.1-Instruct (8B)2026.0583.3281.4516.8672.15
w/ GRPOBackbone=Qwen-3-Instruct (8B)2026.0583.0980.8620.2260.77
w/ GRPOBackbone=LLaMa-3.1-Instruct (8B)2026.0582.5379.9720.2260.77
w/ GRPO*Backbone=Qwen-3-Instruct (8B)2026.0581.8879.9216.8171.26
w/ SFTBackbone=LLaMa-3.1-Instruct (8B)2026.0581.7879.7823.9761.74
w/ SFTBackbone=Qwen-3-Instruct (8B)2026.0581.7176.7419.5762.7
w/ GRPO*Backbone=LLaMa-3.1-Instruct (8B)2026.0581.7176.7418.1866.98
w/ SFT*Backbone=LLaMa-3.1-Instruct (8B)2026.0580.8677.2919.3466.43
w/ SFT*Backbone=Qwen-3-Instruct (8B)2026.0580.4478.9217.2767.12
few-shotBackbone=Qwen-3-Instruct (8B)2026.0570.4569.9926.4957.72
few-shotBackbone=LLaMa-3.1-Instruct (8B)2026.0568.9968.5527.4655.49
zero-shotBackbone=LLaMa-3.1-Instruct (8B)2026.0568.367.5331.6250.34
zero-shotBackbone=Qwen-3-Instruct (8B)2026.0567.4867.3326.9457.66