Web Interaction on GRAPH2EVAL BENCH benchmark leaderboard · SOTA2 Research

Web Interaction on GRAPH2EVAL BENCH

48.89Basic Navigation ScoreGemini 2.5 Flash

Evaluation Results

Method	Links
Gemini 2.5 Flash2025.10		48.89	100	66.67	63.4	100	55	62.5	69.2
GPT-4o mini2025.10		33.34	0	20	37.63	50	17.5	25	33.12
Qwen2.5-VL-72B2025.10		33.34	100	26.67	42.78	83.33	15	56.25	38.8
Gemini 2.5 Flash2025.10		17.78	0	20	11.34	0	25	18.75	14.51
Qwen2.5-VL-32B2025.10		17.78	0	20	15.46	33.33	7.5	12.5	15.14
Qwen2.5-VL-72B2025.10		13.33	100	20	10.31	16.66	27.5	12.5	13.88
Qwen2.5-VL-7B2025.10		11.12	0	6.66	13.92	16.67	7.5	0	11.67
UI-TARS-1.5-7B2025.10		8.89	0	6.67	9.79	0	12.5	12.5	7.19
GPT-4o mini2025.10		6.67	0	0	12.37	0	7.5	0	9.46
UI-TARS-1.5-7B2025.10		6.67	0	0	7.73	0	7.5	12.5	4.91
Qwen2.5-VL-32B2025.10		2.22	100	6.67	1.55	0	7.5	0	2.83
Qwen2.5-VL-7B2025.10		0	0	0	1.03	0	0	0	0.63