Datasets:
docs(README): add configs
Browse files
README.md
CHANGED
|
@@ -16,6 +16,31 @@ size_categories:
|
|
| 16 |
- 1K<n<10K
|
| 17 |
source_datasets:
|
| 18 |
- "MS COCO-2017"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
---
|
| 20 |
|
| 21 |
[SPHERE (Spatial Perception and Hierarchical Evaluation of REasoning)](https://arxiv.org/pdf/2412.12693) is a benchmark for assessing spatial reasoning in vision-language models. It introduces a hierarchical evaluation framework with a human-annotated dataset, testing models on tasks ranging from basic spatial understanding to complex multi-skill reasoning. SPHERE poses significant challenges for both state-of-the-art open-source and proprietary models, revealing critical gaps in spatial cognition.
|
|
|
|
| 16 |
- 1K<n<10K
|
| 17 |
source_datasets:
|
| 18 |
- "MS COCO-2017"
|
| 19 |
+
configs:
|
| 20 |
+
- config_name: distance_and_counting
|
| 21 |
+
data_files: "coco_test2017_annotations/combine_2_skill/distance_and_counting.json"
|
| 22 |
+
- config_name: distance_and_size
|
| 23 |
+
data_files: "coco_test2017_annotations/combine_2_skill/distance_and_size.json"
|
| 24 |
+
- config_name: position_and_counting
|
| 25 |
+
data_files: "coco_test2017_annotations/combine_2_skill/position_and_counting.json"
|
| 26 |
+
- config_name: object_manipulation
|
| 27 |
+
data_files: "coco_test2017_annotations/reasoning/object_manipulation.json"
|
| 28 |
+
- config_name: object_manipulation_w_intermediate
|
| 29 |
+
data_files: "coco_test2017_annotations/reasoning/object_manipulation_w_intermediate.json"
|
| 30 |
+
- config_name: object_occlusion
|
| 31 |
+
data_files: "coco_test2017_annotations/reasoning/object_occlusion.json"
|
| 32 |
+
- config_name: object_occlusion_w_intermediate
|
| 33 |
+
data_files: "coco_test2017_annotations/reasoning/object_occlusion_w_intermediate.json"
|
| 34 |
+
- config_name: counting_only
|
| 35 |
+
data-files:
|
| 36 |
+
- "coco_test2017_annotations/single_skill/counting_only-paired-distance_and_counting.json"
|
| 37 |
+
- "coco_test2017_annotations/single_skill/counting_only-paired-position_and_counting.json"
|
| 38 |
+
- config_name: distance_only
|
| 39 |
+
data_files: "coco_test2017_annotations/single_skill/distance_only.json"
|
| 40 |
+
- config_name: position_only
|
| 41 |
+
data_files: "coco_test2017_annotations/single_skill/position_only.json"
|
| 42 |
+
- config_name: distance_only
|
| 43 |
+
data_files: "coco_test2017_annotations/single_skill/size_only.json"
|
| 44 |
---
|
| 45 |
|
| 46 |
[SPHERE (Spatial Perception and Hierarchical Evaluation of REasoning)](https://arxiv.org/pdf/2412.12693) is a benchmark for assessing spatial reasoning in vision-language models. It introduces a hierarchical evaluation framework with a human-annotated dataset, testing models on tasks ranging from basic spatial understanding to complex multi-skill reasoning. SPHERE poses significant challenges for both state-of-the-art open-source and proprietary models, revealing critical gaps in spatial cognition.
|