Text-to-Image Generation on R2I-Bench benchmark leaderboard · SOTA2 Research

Text-to-Image Generation on R2I-Bench

71Causal Accuracygpt-image-1

Evaluation Results

Method	Links
gpt-image-12025.11		71	81	83	87	88	83
DALL-E-32025.11		64	69	78	76	69	73
ImAgent2025.11		53	62	58	68	71	62
Self-Rewrite2025.11		47	54	54	67	69	58
AFS-Search2026.03		45	58	51	66	20	48
LLM4GENSD1.52025.11		45	55	55	48	39	51
ImAgent2025.11		44	52	47	58	54	51
EMU32026.03		41	61	44	62	9	43
EMU32025.11		41	55	44	59	61	52
Lumina-Image 2.02026.03		40	56	49	65	13	45
Lumina-Image 2.02025.11		40	56	49	65	43	52
Vanilla2025.11		40	53	47	65	68	54
Janus-Pro-7B2026.03		36	46	45	64	7	40
Vanilla2025.11		36	46	45	60	46	47
FLUX.1-dev2026.03		35	37	39	48	5	33
Omnigen2026.03		34	51	43	60	18	41
Omnigen2025.11		34	51	43	60	47	48
Show-o2026.03		30	57	42	56	12	39
Show-o2025.11		30	42	42	59	57	46
ELLASD1.52025.11		29	40	40	44	32	39
Self-Rewrite2025.11		28	35	40	48	39	38
Sana-1.52026.03		21	49	49	67	13	40
Sana-1.52025.11		21	49	49	67	48	49
Lumina-T2I2026.03		18	38	38	49	13	31
SD3-medium2025.11		18	55	54	64	50	53
Lumina-T2I2025.11		18	38	38	49	45	39
LlamaGen2026.03		12	35	38	49	7	28
LlamaGen2025.11		12	38	38	39	35	36