List your company

Helpful and Harmless Response Generation on PKU-SafeRLHF alpaca2-7b (test)

8.53HelpfulnessDPO-bettersafe
1.41643.26325.116.9568Apr 19, 2026

Evaluation results

MethodLinks
DPO-bettersafeBackbone=Qwen3-4B2026.048.539.6898.698.2
Cat-DPOBackbone=Qwen3-4B2026.048.479.7399.498.7
Cat-DPOBackbone=Alpaca-7B2026.048.389.4494.997.2
DPO-harmlessBackbone=Qwen3-4B2026.048.259.7399.699.3
SafeDPOBackbone=Alpaca-7B2026.048.29.2693.796.2
SafeDPOBackbone=Qwen3-4B2026.048.059.2996.297.3
SACPOBackbone=Qwen3-4B2026.047.929.6599.499.7
DPO-harmlessBackbone=Alpaca-7B2026.047.768.8388.793.7
DPO-bettersafeBackbone=Alpaca-7B2026.047.698.728690.5
SACPOBackbone=Alpaca-7B2026.047.658.9392.495.1
DPO-helpfulBackbone=Qwen3-4B2026.047.619.0593.595.2
DPO-helpfulBackbone=Alpaca-7B2026.046.537.3869.876.1
BaseBackbone=Qwen3-4B2026.045.155.9455.253.1
BaseBackbone=Alpaca-7B2026.041.691.99.110.4