README.md

metadata

task_categories:
  - visual-question-answering
language:
  - en

VLMsAreBlind Benchmark

refactor VLMsAreBlind Benchmark to add support for VLMEvalKit.

Benchmark Information

1. Number of questions: 8016
2. Question type: Free-form + Y/N
3. Question format: image + text
4. Answer type: integer, (x, y) tuple, Y/N, single letter

Reference

• VLMEvalKit
• VLMsAreBlind Benchmark