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

Multimodal QA on LLaVABench (test)

48LLaVABench Score

PivotMerge

Updated 2mo ago

Evaluation Results

Method	Links
PivotMerge 2026.04		48
TIES w/ DARE 2026.04		41.5
CC12M Split 3 2026.04		41.3
Weight Average 2026.04		41
MetaGPT 2026.04		41
TIES Merging 2026.04		40.5
CC12M Split 5 2026.04		37.4
Task Arithmetic 2026.04		37.1
TSV-M 2026.04		36.9
CC12M Split 1 2026.04		33.1
CC12M Split 4 2026.04		33
CC12M Split 2 2026.04		28.2
Iso-C 2026.04		17.6