List your company

Sycophancy Evaluation on SycophancyEval

54.2Sycophancy RateLag-DPO
52.46464.18275.987.618Apr 1, 2026

Evaluation results

MethodLinks
Lag-DPOBase Model=Qwen3-4B-Instruct-25072026.0454.245.8
VanillaBase Model=Qwen3-4B-Instruct-25072026.0487.212.9
SFT-CombinedBase Model=Qwen3-4B-Instruct-25072026.0487.612.5
SafeRLHFBase Model=Qwen3-4B-Instruct-25072026.0487.812.2
Multi-Neg DPOBase Model=Qwen3-4B-Instruct-25072026.0489.410.6
PCGrad-DPOBase Model=Qwen3-4B-Instruct-25072026.0489.610.4
MODPOBase Model=Qwen3-4B-Instruct-25072026.0496.23.8
SACPOBase Model=Qwen3-4B-Instruct-25072026.0497.42.6
SFT-AnchorBase Model=Qwen3-4B-Instruct-25072026.0497.62.4