Share your thoughts, 1 month free Claude Pro on usSee more

Object Hallucination Assessment on CHAIR (test)

58.2CS (%)

DoLa

Updated 18d ago

Evaluation Results

Method	Links
DoLa 2025.04		58.2	16.9	79.4	96.35	-	-	-	-	-	-	-	-	-	-	-	-
Fuyu 2025.08		57.3	18	-	-	17.7	60.3	17.1	60.7	24.7	61.3	17.7	55	19.8	57.7	10.8	48.7
VCD 2025.04		55.2	16.8	75.4	92.89	-	-	-	-	-	-	-	-	-	-	-	-
Beam search 2025.04		51.2	13.9	78.2	91.65	-	-	-	-	-	-	-	-	-	-	-	-
Beam search 2025.04		49.6	13.8	76.7	93.01	-	-	-	-	-	-	-	-	-	-	-	-
DoLa 2025.04		47.2	13.6	77.8	88.23	-	-	-	-	-	-	-	-	-	-	-	-
OPERA 2025.04		47.2	13.5	78.3	87.85	-	-	-	-	-	-	-	-	-	-	-	-
AGLA 2025.04		46.6	13.4	78.5	91.69	-	-	-	-	-	-	-	-	-	-	-	-
Greedy 2025.04		45.4	13.4	77.5	89.09	-	-	-	-	-	-	-	-	-	-	-	-
mPLUG-Owl2 2025.08		37.2	12.4	-	-	13.5	40.7	11.6	37	17.2	44.7	13.9	38.7	10.4	35	7.9	27.3
LLaVA-1.5 2025.08		35.3	10.3	-	-	12	43.3	9.1	31.7	11.4	37.7	11.2	34.7	11.4	37.7	6.8	26.7
InstructBLIP 2025.08		35.2	10.1	-	-	13.7	45	9.4	36.7	12	37.3	8.9	31.7	10.2	35.7	6.6	25
LLaVA-1.5-13b 2025.08		33.9	9.8	-	-	10.7	38.7	10.3	35	11	37.3	9.6	33.3	10	31.7	7.4	27.3
MiniGPT-4 2025.08		33.9	10.6	-	-	12	38.7	9.8	33	11.3	34.7	10.6	32.7	11.3	33.3	8.8	31
Qwen-VL 2025.08		33	10.3	-	-	13.2	42.7	8.3	30.3	14.8	39.3	10.2	30.7	10.2	35	5	19.7
TARAC 2025.04		30	8.1	72	83.24	-	-	-	-	-	-	-	-	-	-	-	-
Idefics2-8b 2025.08		24.9	9.2	-	-	9.5	23.7	8.6	23.3	9.9	24.7	11.4	31.7	9.2	26	6.5	20
Phi3V 2025.08		24.7	9.8	-	-	11.3	27.3	9.8	25.7	11.3	20.7	9.6	26.7	9.8	25.7	6.8	22.3
LLaVA-v1.6-mistral-7b 2025.08		23.9	9.4	-	-	10.9	25.7	8.8	22	9.9	24	10.4	26.7	10.1	25.7	6	19.3
Tinyllava 2025.08		23.6	9.5	-	-	13.6	34	8.1	20.3	9.3	22	10.4	24	9.2	22.3	6.5	19
GPT4o 2025.08		21.3	7.1	-	-	10.3	29	7.5	23	7.9	21.3	6.7	20	5.6	17.7	4.7	16.7
Gemini-1.5-PRO 2025.08		18.2	9.8	-	-	13.5	25.3	7.9	14.3	13.1	18.7	11.3	19.3	7	13.3	6.2	18.3