The Role of Chain-of-Thought in Complex Vision-Language Reasoning Task
2023-11-15Unverified0· sign in to hype
Yifan Wu, Pengchuan Zhang, Wenhan Xiong, Barlas Oguz, James C. Gee, Yixin Nie
Unverified — Be the first to reproduce this paper.
ReproduceAbstract
The study explores the effectiveness of the Chain-of-Thought approach, known for its proficiency in language tasks by breaking them down into sub-tasks and intermediate steps, in improving vision-language tasks that demand sophisticated perception and reasoning. We present the "Description then Decision" strategy, which is inspired by how humans process signals. This strategy significantly improves probing task performance by 50%, establishing the groundwork for future research on reasoning paradigms in complex vision-language tasks.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| Winoground | GPT-4V (CoT, pick b/w two options) | Text Score | 75.25 | — | Unverified |
| Winoground | GPT-4V (pick b/w two options) | Text Score | 69.25 | — | Unverified |