Multimodal Embodied Reasoning on BEAR 1.0 (test) benchmark leaderboard · SOTA2 Research

Multimodal Embodied Reasoning on BEAR 1.0 (test)

87.5Task Process Reasoning (PRG)Human

Evaluation Results

Method	Links
Human2025.10		87.5	95.5	92	93.5	83	77	82	92	96.5	94	89	94.5	83.5	88.5	92.5	89.4
GPT-52025.10		59.67	70	63.69	54.9	41.1	32.6	35.2	61	66.99	67.34	49.67	72.31	50.24	47	40	52.17
Gemini-2.5-Pro2025.10		52	55	42.48	55.41	14.4	10.3	17.7	49	66.67	65.99	48.33	64.5	40.1	44	31.43	41.46
Gemini-2.5-Flash2025.10		48.33	46.76	33.33	39.49	18.3	14.5	15.6	43.67	64.42	63.97	45	61.24	43	44.67	31.43	38.24
Claude-4-Sonnet2025.10		44	39.12	40.86	45.54	22.1	17.3	19.7	37.67	50	49.16	38	46.25	42.51	39.67	17.14	33.05
GPT-4o2025.10		43.67	40.59	27.12	34.39	22.7	11.8	20.2	46	41.99	35.35	30.67	60.91	33.33	31	31.43	32.9
InternVL3-8B2025.10		43	52.65	42.48	43.95	36.9	27.5	29.7	33.67	51.28	46.8	27.67	50.16	32.37	20	8.57	33.32
Qwen2.5-VL-32B-Instruct2025.10		42.67	27.35	27.78	42.68	2	1.8	1.7	42.33	55.45	52.19	26.67	47.23	26.57	22.67	20	28.33
InternVL2-26B2025.10		41.33	21.18	15.36	18.79	20.1	20.2	14.7	34.33	53.21	43.77	30.33	26.06	26.57	22	11.29	25.66
InternVL3-14B2025.10		41	37.94	27.78	32.8	30.4	25.8	27.6	33	51.28	49.49	31.43	43	28.02	21.33	28.57	33.93
Qwen2.5-VL-7B-Instruct2025.10		40.67	6.18	1.63	0.96	0.7	0.3	0.9	32.33	54.49	48.15	30	38.44	31.4	21	22.86	21.44
InternVL2-40B2025.10		40	23.24	21.24	22.29	32.9	26.9	26.8	33.67	57.69	41.75	28	40.39	29.47	18.67	11.43	28.38
DeepSeek-VL-7B2025.10		37.67	14.12	8.5	9.24	27.6	16	23.1	27.33	41.03	38.72	22.67	42.02	37.68	32	20	23.89
Molmo-7B-D-09242025.10		37.67	23.53	19.28	25.48	10.9	8.2	10.9	31	45.51	41.41	23.33	49.84	29.47	26	5.71	24.22
LLava-NeXT-Interleave-7B2025.10		37.33	6.47	3.59	2.55	0	0	0	26	37.18	37.04	20.67	37.79	27.54	19.67	5.71	14.64
LLaVa-NeXT-Llama3-8B2025.10		36.67	2.94	1.31	0.96	32	24.6	20.5	29.67	39.42	37.71	23	40.39	33.82	24	14.29	21.65
Random Choice2025.10		25	—	—	—	—	—	—	25	25	25	25	25	28	25	25	—