Zhong
commited on
Update README.md
Browse files
README.md
CHANGED
|
@@ -155,17 +155,3 @@ Model performance under different **Instr.** (Instruction) settings: **Dft.** (D
|
|
| 155 |
The `RefOI` dataset is designed for fine-grained REG/REC analysis. It distinguishes between **COCO** and **non-COCO classes**, and between scenes with **single presence vs. co-occurrence** of the same class.
|
| 156 |
We encourage users to leverage these distinctions for deeper insights and invite community contributions to expand non-COCO annotations.
|
| 157 |
|
| 158 |
-
|
| 159 |
-
## Citation
|
| 160 |
-
If you find our data useful, please consider citing our work:
|
| 161 |
-
|
| 162 |
-
```bibtex
|
| 163 |
-
@misc{vlmreg2025,
|
| 164 |
-
title = {Vision–Language Models Are Not Pragmatically Competent in Referring Expression Generation},
|
| 165 |
-
year = {2025},
|
| 166 |
-
note = {Under review at COLM 2025, ACL GEM 2025},
|
| 167 |
-
url = {https://vlm-reg.github.io},
|
| 168 |
-
}
|
| 169 |
-
```
|
| 170 |
-
|
| 171 |
-
|
|
|
|
| 155 |
The `RefOI` dataset is designed for fine-grained REG/REC analysis. It distinguishes between **COCO** and **non-COCO classes**, and between scenes with **single presence vs. co-occurrence** of the same class.
|
| 156 |
We encourage users to leverage these distinctions for deeper insights and invite community contributions to expand non-COCO annotations.
|
| 157 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|