Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Visual Instruction Following research benchmarks · SOTA2 Research

Visual Instruction Following

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

LLaVA-Bench Wild

102.3Score

71

May 19, 2026

FronTalk Multi-Turn 1.0 (test)

75PR Score

32

Feb 26, 2026

102Score

28

Feb 26, 2026

FronTalk Single-Turn 1.0 (test)

Claude-4-Sonnet

77.5PR

20

Feb 26, 2026

47.8Accuracy

16

Jun 19, 2026

SUN Instructive Visual Benchmark

44.68Accuracy

16

Jun 19, 2026

Caltech 101 Instructive Visual Benchmark

38.74Accuracy

16

Jun 19, 2026

ImageNet Instructive Visual Benchmark

85Accuracy

16

Jun 19, 2026

79.1Overall Score

15

Mar 31, 2026

Visual Instruction Total (test)

153.04Avg. Response Length (Words)

6

Feb 26, 2026

LLaVA-Bench 100 images (test)

6.53Accuracy

6

Feb 26, 2026

Visual Instruction Out-Of-Distribution - Hard (test)

BeautifulPrompt

88Win Ratio (Human)

5

Feb 26, 2026

Visual Instruction Out-Of-Distribution - Simple (test)

BeautifulPrompt

93Human Win Ratio

5

Feb 26, 2026

Visual Instruction Out-Of-Distribution (test)

BeautifulPrompt

90Human Win Ratio

5

Feb 26, 2026

Visual Instruction In-Distribution - Hard (test)

BeautifulPrompt

89Human Win Ratio

5

Feb 26, 2026

Visual Instruction In-Distribution - Simple (test)

BeautifulPrompt

89Human Win Ratio

5

Feb 26, 2026

Visual Instruction In-Distribution (test)

BeautifulPrompt

89Human Win Ratio

5

Feb 26, 2026

Popular datasets

Visual Instruction LLaVA-Bench FronTalk ImageNet RefCOCO

Follow for updates

Join SOTA2 Contact