Instruction-guided image editing

Benchmarks

Dataset Name	SOTA Method	Metric
GEdit-Bench EN Full set	Ours	G_SC8.44	33	1mo ago
KRIS-Bench		Attribute Accuracy85.61	27	1mo ago
MagicBrush single-turn (test)	MagicBrush	CLIP Similarity (Image)0.9332	13	3mo ago
EmuEdit	I2E	DINO Score0.9404	10	15d ago
ZONE (test)	WYS	CLIP-T0.301	7	3mo ago
MagicBrush multi-turn (test)	WYS	CLIP-T0.313	7	3mo ago
HumanEdit (test)	Bagel-GVCoT	CLIP-I0.9451	6	3mo ago
I2E-BENCH 1.0 (test)	IP2P	LPIPS-U0.2137	6	3mo ago
MagicBrush	I2E	LPIPS-U0.0446	6	3mo ago
InstructBrush User Study (randomly sampled pairs)	Instruct-CLIP	Total Samples566	6	3mo ago
Human Evaluation User Study (test)	ZONE	Success Rate (SR)69.4	6	3mo ago
Real-world local, global, and compound editing evaluation sets (test)	HorizonWeaver	Preference (%)61.1	5	1mo ago
Gedit-Bench	BAGEL	Background Change Score7.06	3	1d ago
HumanEdit Relation	MLLM-as-a-Judge	Overall Average Score6.153	3	3mo ago
HumanEdit (All Edits)	MLLM-as-a-Judge	Image Preservation6.074	2	3mo ago
HumanEdit Counting	MLLM-as-a-Judge	Image Preservation Score5.2	2	3mo ago
HumanEdit Action	MLLM-as-a-Judge	Image Preservation Score6.203	2	3mo ago
HumanEdit Replace	MLLM-as-a-Judge	Image Preservation Score6.352	2	3mo ago
HumanEdit Remove	MLLM-as-a-Judge	Image Preservation Score5.971	2	3mo ago
HumanEdit Add	MLLM-as-a-Judge	Image Preservation6.555	2	3mo ago
GEdit-Bench (test)	BAGEL-NHR-EDIT	SC7.8	2	3mo ago
ImgEdit-Bench (test)	BAGEL-NHR-EDIT	Overall Score3.33	2	3mo ago
HumanEdit Counting		Human Avg Score66.8	1	3mo ago
HumanEdit Replace		Human Avg80.1	1	3mo ago
HumanEdit Add 1.0		Human Avg0.786	1	3mo ago

Showing 25 of 25 rows