ObjectQA on Q-Real benchmark leaderboard · SOTA2 Research

ObjectQA on Q-Real

76.2Naturalness Q1 AccuracyInternVL2.5-8B*

Updated 4mo ago

Evaluation Results

Method	Links
InternVL2.5-8B*2025.11		76.2	74	75.8	80.7	75.2	73	75.3	81.4	76.7	69.1	65.8	84.2	70.7	70.7	73.1	76	74.4	74.4	75.2	81.9	94.5	79.8	72.7	93.5
Qwen3-VL-8B-Instruct*2025.11		76.2	72.8	76.5	84.3	75.5	72.4	76.1	84	77.7	71.8	66.3	84.3	70.9	71.9	72.7	78.7	73.6	74.7	73.9	82.1	94.6	80.1	73.7	95.3
GPT-4o2025.11		75.2	68.2	73.9	—	70.6	60.2	73.5	—	69.3	62.3	63.2	—	60.3	59.1	63.2	—	60.5	72.8	44.1	—	92.9	70	68.8	—
Qwen2.5-VL-7B-Instruct*2025.11		75.1	71.6	75.7	82.8	74.3	70	75.7	82.6	77	68.5	66.9	82.8	70.2	71.9	71.7	77.2	73.5	74.4	73.8	81.2	93.3	68.9	70.1	92.9
InternVL3-8B*2025.11		73.7	69.2	65.5	81	73.4	70	64.6	80.5	79.4	75.2	76.7	87.4	70.1	70.6	72.3	78.4	72.7	72.4	73.7	82.2	93.9	72.9	71.5	95.8
Qwen3-VL-8B-Instruct2025.11		70.2	69.4	68.6	77.6	66.1	65.1	65.2	72.4	56.2	28.3	19.5	39.8	60	59	67.6	66.5	62.2	64.2	61.2	66.7	92.8	87.4	55.9	91.5
InternVL2.5-8B2025.11		70	69.2	68.7	28.5	65.1	60.2	69.9	34.1	65.6	57.4	22.1	68.3	60.8	60.4	66.8	55.4	55.4	53.6	68	40.2	89.6	64.8	22.3	88.6
InternVL3-8B2025.11		68.2	69	65.3	77.6	64.3	59.1	70.2	74.8	65.9	59.6	22.7	71.4	58.9	68.9	50.7	67.7	62.4	71.8	43.4	74.8	89.8	80.4	18.5	93.3
Q-Eval-Score2025.11		67.5	72.3	61.3	74.1	64.5	63.6	63.4	68.8	68.3	55.5	55.2	71.1	60.9	58.4	70.6	66.6	58.9	55.8	69.8	69.2	85.2	40.6	52	87.3
Qwen2.5-VL-7B-Instruct2025.11		60.5	76.9	38.8	77.6	63	73.4	49.7	75	65.2	71.4	8.2	71.1	56.3	81	35.1	69	67.3	82.8	58.7	75.4	89.2	73.6	7.4	88.1