Sycophancy Evaluation on VISE benchmark leaderboard · SOTA2 Research

Sycophancy Evaluation on VISE

58.04Strong BiasGemini-1.5-Pro

Evaluation Results

Method	Links
Gemini-1.5-Pro2025.06		58.04	33.96	47.94	42.05	41.83	19.59	22.39	58.04	37.97
Qwen2.5-VL2025.06		57.66	38.16	43.41	45.32	60.54	30.55	38.79	60.54	44.92
LLaVA-Onevision2025.06		54.39	54.51	55.34	59.55	57.05	57.1	26.82	59.55	52.11
InternVL 2.52025.06		33.83	26.45	22.46	16.69	40.45	41.44	30.41	41.44	30.25
Qwen2.5-VL2025.06		28.34	16.23	17.81	13.34	17.53	4.77	34.56	34.56	18.94
Qwen2.5-VL2025.06		26.85	11.87	21.9	17.25	10.29	8.39	10.29	26.85	15.26
InternVL 2.52025.06		25.75	21.48	16.01	13.66	25.66	19.51	25.07	25.75	21.02
GPT 4o mini2025.06		8.72	7.72	9.53	6.76	11.76	6.69	45.96	45.96	13.88
VideoChat-Flash2025.06		7.55	5.09	4.16	2.67	13.36	52.68	24.39	52.68	15.7