List your company

Sycophancy on FaithfulQA

87.58Sycophancy ScoreREP
45.491256.418167.34578.2719Feb 7, 2026

Evaluation results

MethodLinks
REPBackbone=Mistral-7B-Instruct-v0.12026.0287.58
CAABackbone=Mistral-7B-Instruct-v0.12026.0286.77
STEER2ADAPTBackbone=Mistral-7B-Instruct-v0.12026.0284.68
STEER2ADAPTBackbone=Llama-3.1-8B-Instruct2026.0284.29
ICLBackbone=Mistral-7B-Instruct-v0.12026.0283.73
Few-ShotBackbone=Llama-3.1-8B-Instruct, n=22026.0282.44
Direct InferenceBackbone=Mistral-7B-Instruct-v0.12026.0281.95
Few-ShotBackbone=Llama-3.1-8B-Instruct, n=12026.0279.49
REPBackbone=Llama-3.1-8B-Instruct2026.0277.68
CAABackbone=Llama-3.1-8B-Instruct2026.0275.37
Direct InferenceBackbone=Llama-3.1-8B-Instruct2026.0272.64
ICLBackbone=Llama-3.1-8B-Instruct2026.0271.7
CAABackbone=Qwen-2.5-7B-Instruct2026.0268.38
Few-ShotBackbone=Qwen-2.5-7B-Instruct, n=22026.0268.33
Few-ShotBackbone=Qwen-2.5-7B-Instruct, n=12026.0267.99
STEER2ADAPTBackbone=Qwen-2.5-7B-Instruct2026.0265.93
ICLBackbone=Qwen-2.5-7B-Instruct2026.0264.25
Few-ShotBackbone=Mistral-7B-Instruct-v0.1, n=12026.0264.1
Direct InferenceBackbone=Qwen-2.5-7B-Instruct2026.0262.66
REPBackbone=Qwen-2.5-7B-Instruct2026.0262.27
Few-ShotBackbone=Mistral-7B-Instruct-v0.1, n=22026.0247.11