Multimodal Image Understanding on LLaVA-1.5-7B Evaluation Suite benchmark leaderboard · SOTA2 Research

Multimodal Image Understanding on LLaVA-1.5-7B Evaluation Suite

61.9GQA ScoreBaseline

Evaluation Results

Method	Links
Baseline2026.07		61.9	64	57.9	1,834.8	86.2	67.9	45.7	54.4	31.5	100	—
TORINO-M2026.07		60.3	62.2	56.4	1,778.8	85.3	67.6	43.6	54.6	33.1	98.7	26.4
TORINO-P2026.07		60.1	62.6	56	1,749.5	85.7	67.2	43.7	54.8	32.4	98.3	27.6
TORINOFB-P2026.07		59.8	62.5	56.8	1,756.5	85.6	67.5	42.7	54.2	32	97.9	28.1
TORINOFB-M2026.07		59.8	63	56.7	1,740.4	85.1	67.4	42.5	53.9	32.7	98	26.6
Folder2026.07		59.4	60.7	52.8	1,732.3	86.4	68.4	39.6	54.2	28.9	95	16.6
Random2026.07		59.1	61.2	54.1	1,772.6	84.7	67.7	34.3	55.4	29.8	94.5	13.6
PruneSID2026.07		59.1	62.1	55.8	1,713.8	85.8	67.3	40.8	54.9	31.3	96.7	64.6
TORINO-P2026.07		58.8	61	55.2	1,763.7	84	68	41.6	56.1	31.6	97.1	29.6
TORINO-M2026.07		58.6	60.8	54.8	1,742.6	83.6	67.8	41.4	55.5	30	96	28.5
TORINOFB-P2026.07		58.5	61.9	55.2	1,745.9	83.8	68	40.3	55.8	30.9	96.4	30
PruMerge2026.07		58.3	62.3	56.4	1,713.7	84.7	67.4	40.8	55.8	30.9	96.7	71.2
TORINOFB-M2026.07		58.2	61.9	54.9	1,715.7	83.2	68.4	40	55.6	28.7	95.2	29.1
PruneSID2026.07		57.9	60.7	52.6	1,724.7	84.8	67.5	40.1	56.1	27.4	94.3	60.1
Random2026.07		57	58.4	49.8	1,673.8	82.2	67.3	28	57	26.9	89.6	13.5
PruMerge2026.07		56.5	61.9	55.3	1,677.5	74.7	67.9	39.3	56.4	32.2	94.9	48.2
Folder2026.07		56.4	56.4	46.8	1,641	82.5	67.1	32.1	56	29.3	90.1	17.2
PruneSID2026.07		56.4	59.6	50.6	1,735	84.4	67.2	37.6	57	26.7	92.8	23
TORINO-P2026.07		55.2	59.5	51.4	1,612.8	77.4	67.7	33.1	56.3	30.1	91.1	21.8
TORINO-M2026.07		54.9	58	50.2	1,560.5	75.2	67.4	32.5	56.7	29.4	89.6	20.7
TORINOFB-P2026.07		54.7	59.2	50.6	1,648.4	77.3	67.9	32.5	56.2	28.8	90.4	22
TORINOFB-M2026.07		54.3	57.1	49.1	1,625.7	75	67.2	31.8	56.2	28	88.7	20.8
Random2026.07		53.9	51.5	39.9	1,575.5	74.1	67.3	20.4	56.2	24.7	81.5	14
PruMerge2026.07		53	59.9	51.1	1,631	70.5	67.2	37.9	56.7	29.4	90.9	25.7
Folder2026.07		51	46.2	32.7	1,283.5	59.6	65.4	22.9	56	24.4	75.2	21