Misinformation Detection on COCO benchmark leaderboard · SOTA2 Research

Misinformation Detection on COCO

81.9AccuracyIT_zs

Evaluation Results

Method	Links
IT_zs2026.01		81.9	80.1	78.5	79.2	67.8
gpt-4.12026.01		81.2	81.7	75.9	75.5	69.6
RAEmoLLM2026.01		81.2	82.2	76.1	75.8	69.6
RAAR-14b2026.01		79.9	79.2	74.8	75.9	71.2
MDFEND2024.06		79.88	79.39	79.88	77.93	—
Deepseek-R2026.01		79.3	81.8	72.6	73.9	70.5
gpt-5-mini2026.01		78.6	80.7	71.4	72.5	68.8
Mistral-7b2024.06		78.14	80.53	78.14	79.31	—
gpt-4.12026.01		78.1	59.6	53.7	53.7	63.9
Deepseek-R2026.01		77.6	81	70.5	70.8	67.9
RAAR-8b2026.01		77.6	77.9	70.7	72.7	68.9
Vicuna-7b2024.06		77.56	79.56	77.56	75.01	—
gpt-5-mini2026.01		76.5	79.8	68.6	69.1	67.3
Deepseek-C2026.01		76.3	79.7	68.4	68.9	68.4
Qwen-14b2026.01		75	75.3	66.6	67.7	65.9
Qwen-32b2026.01		73.5	75	65.2	66.6	66.2
BERT2024.06		72.88	75.1	72.88	63.56	—
MDFEND2026.01		72	72.2	64	66.2	57
Qwen-8b2026.01		71.6	71.2	62.6	63.4	58.6
EDDFN2026.01		71.5	68.1	68	68	54.5
Deepseek-C2026.01		71.2	74.7	60.9	61.2	64.6
Qwen-14b2026.01		70.6	72.6	60.4	61	57
Qwen-32b2026.01		68.7	72.9	57.5	56.6	61.4
IT_ra2026.01		66	55.2	57.2	53	53.2
Qwen-8b2026.01		64.1	66.6	52.9	51.6	55
CANMD2026.01		62.8	59.5	58.3	57.9	46.8
Llama-8b2026.01		59.8	59.2	58.6	56.8	47.2
MetaAdapt2026.01		51.1	49.2	50.4	48.7	53.2
Llama-8b2026.01		47.1	46.3	34	35.8	38.5
MOSE2026.01		46.6	35.5	34.3	34.8	40.4