Datasets:
Link SnapBench paper
Browse files
README.md
CHANGED
|
@@ -20,6 +20,8 @@ configs:
|
|
| 20 |
|
| 21 |
Source repository: <https://github.com/zrchen03/SnapBench>
|
| 22 |
|
|
|
|
|
|
|
| 23 |
# SnapBench: Benchmarking Snap-and-Ask Multimodal Retrieval for Mobile Interactions
|
| 24 |
|
| 25 |
SnapBench is a benchmark for **snap-and-ask** mobile interactions: a user captures a photo and asks a short English question. Each query pairs an image with text; the gallery contains image–caption pairs to retrieve from. The benchmark includes a clean split plus text and image perturbations to simulate real-world query degradation.
|
|
|
|
| 20 |
|
| 21 |
Source repository: <https://github.com/zrchen03/SnapBench>
|
| 22 |
|
| 23 |
+
Paper: [SnapBench: Benchmarking Snap-and-Ask Multimodal Retrieval for Mobile Interactions](https://arxiv.org/abs/2608.29607)
|
| 24 |
+
|
| 25 |
# SnapBench: Benchmarking Snap-and-Ask Multimodal Retrieval for Mobile Interactions
|
| 26 |
|
| 27 |
SnapBench is a benchmark for **snap-and-ask** mobile interactions: a user captures a photo and asks a short English question. Each query pairs an image with text; the gallery contains image–caption pairs to retrieve from. The benchmark includes a clean split plus text and image perturbations to simulate real-world query degradation.
|