Visual Question Answering (VQA) on VQA v2 val

Metric: Accuracy (higher is better)

LeaderboardDataset

Loading chart...

Results

Sort:

#	Model↕	Accuracy▼	Extra Data	Paper	Date↕	Code
1	BLIP-2 ViT-G OPT 6.7B (fine-tuned)	82.19	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
2	BLIP-2 ViT-G OPT 2.7B (fine-tuned)	81.59	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
3	BLIP-2 ViT-G FlanT5 XL (fine-tuned)	81.55	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
4	BLIP-2 ViT-G FlanT5 XXL (zero-shot)	65.2	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
5	PNP-VQA	63.3	No	Plug-and-Play VQA: Zero-shot VQA by Conjoining L...	2022-10-17	Code
6	BLIP-2 ViT-G FlanT5 XL (zero-shot)	63.1	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
7	BLIP-2 ViT-L FlanT5 XL (zero-shot)	62.6	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
8	LocVLM-L	55.9	No	Learning to Localize Objects Improves Spatial Re...	2024-04-11	Code
9	BLIP-2 ViT-G OPT 6.7B (zero-shot)	54.3	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
10	BLIP-2 ViT-G OPT 2.7B (zero-shot)	53.5	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
11	BLIP-2 ViT-L OPT 2.7B (zero-shot)	50.1	No	BLIP-2: Bootstrapping Language-Image Pre-trainin...	2023-01-30	Code
12	Few VLM (zero-shot)	47.7	No	A Good Prompt Is Worth Millions of Parameters: L...	2021-10-16	Code
13	MetaLM	41.1	No	Language Models are General-Purpose Interfaces	2022-06-13	Code
14	VLKD(ViT-B/16)	38.6	No	-	-	-
15	Frozen	29.5	No	Multimodal Few-Shot Learning with Frozen Languag...	2021-06-25	-

#1BLIP-2 ViT-G OPT 6.7B (fine-tuned)SOTA
82.19
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#2BLIP-2 ViT-G OPT 2.7B (fine-tuned)
81.59
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#3BLIP-2 ViT-G FlanT5 XL (fine-tuned)
81.55
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#4BLIP-2 ViT-G FlanT5 XXL (zero-shot)
65.2
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#5PNP-VQASOTA
63.3
Accuracy· 2022-10-17
Plug-and-Play VQA: Zero-shot VQA by Conjoining Large Pretrained Models with Zero Training Code
#6BLIP-2 ViT-G FlanT5 XL (zero-shot)
63.1
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#7BLIP-2 ViT-L FlanT5 XL (zero-shot)
62.6
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#8LocVLM-L
55.9
Accuracy· 2024-04-11
Learning to Localize Objects Improves Spatial Reasoning in Visual-LLMs Code
#9BLIP-2 ViT-G OPT 6.7B (zero-shot)
54.3
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#10BLIP-2 ViT-G OPT 2.7B (zero-shot)
53.5
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#11BLIP-2 ViT-L OPT 2.7B (zero-shot)
50.1
Accuracy· 2023-01-30
BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models Code
#12Few VLM (zero-shot)SOTA
47.7
Accuracy· 2021-10-16
A Good Prompt Is Worth Millions of Parameters: Low-resource Prompt-based Learning for Vision-Language Models Code
#13MetaLM
41.1
Accuracy· 2022-06-13
Language Models are General-Purpose Interfaces Code
#14VLKD(ViT-B/16)
38.6
Accuracy
No paper
#15FrozenSOTA
29.5
Accuracy· 2021-06-25
Multimodal Few-Shot Learning with Frozen Language Models