Trigger Detection on ProAct-75 1.0 benchmark leaderboard · SOTA2 Research

Trigger Detection on ProAct-75 1.0

62.9mAccProAct-Helper

Evaluation Results

Method	Links
ProAct-Helper2026.02		62.9	76.56	80.08	85.65
ProAct-Helper2026.02		62.36	76.09	79.85	85.57
ProAct-Helper2026.02		61.5	75.4	79.24	85.12
ProAct-Helper2026.02		61.5	75.38	79.32	85.23
Qwen3-VL-Flash2026.02		54.73	70.35	72.06	77.48
Qwen3-VL-30B-A3B-Instruct2026.02		51.61	66.25	73.5	81.89
Qwen3-VL-Plus2026.02		51.52	67.39	69.89	76.41
Gemini-2.5-Flash2026.02		49.73	64.24	72.52	81.44
Qwen3-Omni-30B2026.02		46.86	63.35	65.07	71.3
Qwen2.5-VL-32B-Instruct2026.02		45.72	62.36	63.77	69.64
Gemini-2.5-Pro2026.02		42.89	55.94	69.31	80.21
Qwen2.5-VL-3B-Instruct2026.02		42.33	54.94	69.55	80.59
GPT-4o2026.02		30.59	46.7	47.11	51.38
Gemini-2.5-Pro2026.02		8.69	12.31	11.99	21.41
ProAct-Helper2026.02		8.25	11.56	19.85	33.13
ProAct-Helper2026.02		7.61	10.67	18.68	31.49
Gemini-2.5-Flash2026.02		6.97	10.12	9.42	17.21
Qwen3-VL-30B-A3B-Instruct2026.02		6.67	9.44	8.53	15.72
ProAct-Helper2026.02		6.46	9.27	17.54	29.85
ProAct-Helper2026.02		6.39	9.7	18.23	30.83
Qwen3-VL-Plus2026.02		5.14	7.38	4.66	8.91
Qwen2.5-VL-32B-Instruct2026.02		4.04	6.04	6.14	11.57
GPT-4o2026.02		4.02	6.05	8.7	16
Qwen3-Omni-30B2026.02		3.72	5.42	4.48	8.58
Qwen3-VL-Flash2026.02		3.7	5.51	2.94	5.72
Qwen2.5-VL-3B-Instruct2026.02		1.46	2.42	3.06	5.95