3D Question Answering on MSQA benchmark leaderboard · SOTA2 Research

3D Question Answering on MSQA

47.9Count AccuracySCENECOT

Evaluation Results

Method	Links
SCENECOT2025.10		47.9	82.1	49.6	47.2	51.6	80.3	55.6
SCENECOT2025.10		47.7	82.1	49.6	47.2	51.6	80.3	55.6
MAG-3DSeed-1.62026.04		44.1	88.5	62.2	33.9	53.5	73.7	54.3
LLaVA-3D2025.10		38.5	89.5	57	49.8	38.5	84.4	54.9
Chat-Scene†2025.10		37.4	92	49	47	58.3	83.7	56.6
Chat-Scene2025.10		37.4	92	49	47	58.3	83.7	56.6
MAG-3DGPT-4o2026.04		37	85.6	66.5	31.2	53.2	79.2	53
GPT-3.52026.04		34.8	74.5	75.8	27.9	43	88	50.7
Loc3R-VLM2026.03		33.1	88.5	61.3	57.6	47.2	83.8	58.6
MSR3D†2025.10		32.7	87.5	53.7	44.3	51.5	72.3	52.8
MSR3D2025.10		32.7	87.5	53.7	44.3	51.5	72.3	52.8
Claude-3.5-Sonnet2026.04		32.6	66.3	69.9	30.1	45.5	83.6	49.7
LEO2025.10		32.5	88.5	58.7	44.2	39.6	81.4	54.8
LEO2026.03		32.5	88.5	58.7	44.2	39.6	81.4	54.8
GPT-4o*2025.10		32.3	79.3	79	37	31.7	91.6	52.3
MSR3D2025.10		32.3	93.1	50	46.5	54.1	75.6	54.2
GPT-4o2026.03		32.3	79.3	79	37	31.7	91.6	52.3
MSR3D2026.03		32.3	93.1	50	46.5	54.1	75.6	54.2
Seed-1.62026.04		31.3	75.1	69.4	36.8	24	77.1	47.9
GPT-4o2026.04		31.2	71.4	75.2	31.5	36.7	88	49.7
LEO†2025.10		29.3	87.5	55.1	46.3	45.6	81.4	52.9
LEO2025.10		29.3	87.5	55.1	46.3	45.6	81.4	52.9
SplatTalk‡2025.10		19.6	60.3	44	35.8	35.5	61.8	41.8
SplatTalk2026.03		19.6	60.3	44	35.8	35.5	61.8	41.8
LEO2026.04		0.8	15.5	11.8	7.3	2.3	15.3	7.8