yuhui-zh15/AutoConverter

Official implementation of "Automated Generation of Challenging Multiple-Choice Questions for Vision Language Model Evaluation" (CVPR 2025)

23
/ 100
Experimental

This tool helps researchers and evaluators of Vision Language Models (VLMs) by transforming existing open-ended visual question-answering (VQA) datasets into a multiple-choice format. It takes images and their original open-ended questions as input, and outputs challenging multiple-choice questions, along with correct and plausible incorrect answers. This makes VLM evaluation more objective and efficient for AI researchers and machine learning engineers.

No commits in the last 6 months.

Use this if you need to create standardized, challenging multiple-choice questions from open-ended VQA datasets to rigorously evaluate different Vision Language Models.

Not ideal if you are looking to generate brand new visual questions or if your primary goal is not VLM evaluation.

AI Evaluation Machine Learning Research Vision Language Models Dataset Generation AI Benchmarking
No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 7 / 25
Maturity 8 / 25
Community 6 / 25

How are scores calculated?

Stars

40

Forks

2

Language

Python

License

Last pushed

May 26, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/yuhui-zh15/AutoConverter"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.