Grounded Reasoning on MM-GCoT benchmark leaderboard · SOTA2 Research

Grounded Reasoning on MM-GCoT

78.9Attribute Accuracy (A-Acc)GR3D-8B

Evaluation Results

Method	Links
GR3D-8B2026.05		78.9	77.3	66.7	85	79.6	70.4	71.1	65.7	66.1	78.3	74.2	67.7
ROVER-LSW-GRPO2026.05		76.7	81	68	95.6	90.3	85.9	71.4	73.9	68.9	81.2	81.7	74.3
ROVER-LSW2026.05		75.6	78	66.7	90.8	90.3	81.1	63.8	71.4	64.8	76.7	79.9	70.8
LLaVA-13B GCoT2026.05		74.7	67.8	58	90.3	72.8	68	64.1	60.8	59.3	76.4	67.1	61.8
Qwen2.5-VL-7B2026.05		73.6	72.5	59.8	87.9	56.3	51.5	57.8	64.1	59.1	73.1	64.3	56.8
LLaVA-7B GCoT2026.05		72.8	66.7	56.1	88.3	61.7	56.9	62.3	61.7	61.3	74.5	63.3	58.1
LLaVA-GCoT-7B2026.05		72.8	66.7	56.1	88.3	61.7	56.9	62.3	61.7	61.3	74.5	63.3	58.1
LLaVA-13B2026.05		69.7	12	9.6	83.5	14.6	15.4	58.7	11.6	14.4	70.6	12.7	13.1
LLaVA-13B2026.05		69.7	11.5	11.7	83.3	6.1	6.6	51.9	14.6	17.2	68.3	10.8	11.8
LLaVA-7B2026.05		68.6	9.2	8.8	83	11.2	11.5	58.4	9.1	9.9	70	9.8	10.1
LLaVA-7B2026.05		68.6	9.2	8.8	83	11.2	11.5	58.4	9.1	9.9	70	9.8	10.1
InternVL2.5-8B2026.05		61.9	51.9	42.6	89.3	36.4	31.5	48.3	43.2	39.4	66.5	43.8	37.8
LLaVA-7B2026.05		59.7	6.3	5.6	82.5	0.5	0.6	35.9	5.5	9.7	59.4	4.1	5.3
LLaVA-7B2026.05		59.7	6.3	5.6	82.5	0.5	0.6	35.9	5.5	9.7	59.4	4.1	5.3
InternVL2.5-8B2026.05		54.5	43.1	40.7	76.9	31.9	26.1	37.1	46.2	31.1	56.2	40.4	32.6
Qwen2.5-VL-7B2026.05		48.8	82.6	45.7	80.6	72.8	62.4	26.7	62.3	32.6	52	72.6	46.9