buraksatar commited on
Commit
beda6f8
·
verified ·
1 Parent(s): 55ec373

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +0 -1
README.md CHANGED
@@ -31,7 +31,6 @@ configs:
31
 
32
  Multimodal vision-language models (VLMs) have made substantial progress in various tasks that require a combined understanding of visual and textual content, particularly in cultural understanding tasks, with the emergence of new cultural datasets.
33
  However, these datasets frequently fall short of providing cultural reasoning while underrepresenting many cultures.
34
-
35
  In this work, we introduce the Seeing Culture Benchmark (SCB), focusing on cultural reasoning with a novel approach that requires VLMs to reason on culturally rich images in two stages: i) selecting the correct visual option with multiple-choice visual question answering (VQA), and ii) segmenting the relevant cultural artifact as evidence of reasoning. Visual options in the first stage are systematically organized into three types: those originating from the same country, those from different countries, or a mixed group.
36
  Notably, all options are derived from a singular category for each type. Progression to the second stage occurs only after a correct visual option is chosen.
37
 
 
31
 
32
  Multimodal vision-language models (VLMs) have made substantial progress in various tasks that require a combined understanding of visual and textual content, particularly in cultural understanding tasks, with the emergence of new cultural datasets.
33
  However, these datasets frequently fall short of providing cultural reasoning while underrepresenting many cultures.
 
34
  In this work, we introduce the Seeing Culture Benchmark (SCB), focusing on cultural reasoning with a novel approach that requires VLMs to reason on culturally rich images in two stages: i) selecting the correct visual option with multiple-choice visual question answering (VQA), and ii) segmenting the relevant cultural artifact as evidence of reasoning. Visual options in the first stage are systematically organized into three types: those originating from the same country, those from different countries, or a mixed group.
35
  Notably, all options are derived from a singular category for each type. Progression to the second stage occurs only after a correct visual option is chosen.
36