Embodied AI reasoning on ERIQ 6K (full) benchmark leaderboard · SOTA2 Research

Embodied AI reasoning on ERIQ 6K (full)

96.67Action UnderstandingOurs-3B

Evaluation Results

Method	Links
Ours-3B2025.12		96.67	85.25	90.5	55.36	73.86	51.6	96.44	83.26	75.45	93.1	85.71	82.72
Gemini-2.5-pro2025.12		89.83	67.37	76.67	81.25	75.05	62.28	91.11	87.67	67.07	90.52	87.62	80.55
Qwen3-VL-8B2025.12		89.17	69.38	66.33	65.18	63.76	33.45	89.78	76.65	52.99	86.21	80	75.53
Cosmos-Reason1-7B2025.12		86.33	54.75	65	53.57	63.37	22.78	84	81.94	58.08	69.83	72.38	67.99
GPT-4o-mini2025.12		84.67	71.63	65.17	71.43	68.91	49.82	89.78	79.74	66.77	77.59	68.57	77.61
RoboBrain2.0-7B2025.12		82.33	56.88	61	51.79	58.22	24.56	84.89	83.26	50.9	76.72	57.14	67.38
InternVL-3.5-8B2025.12		78.33	55	59.5	59.82	58.42	27.4	80.89	83.26	50.6	74.14	65.71	66.72
Qwen2.5-VL-7B2025.12		76.83	62.62	60.67	54.46	60.4	24.2	73.78	77.53	50.3	72.41	64.76	66.69
Claude-Sonnet-42025.12		67.67	61.5	61.33	62.5	55.84	32.03	92.89	83.26	37.13	84.48	83.81	65.66
Qwen2.5-VL-3B2025.12		65.5	52.75	55.67	46.43	57.62	25.98	86.22	78.85	34.43	73.28	59.05	58.64