Upload folder using huggingface_hub

README.md

@@ -8,15 +8,60 @@ configs:
 
 # SYNCR
 
-A synthetic cross-video reasoning dataset spanning 4 categories and 8 tasks.
-
-This version contains annotations only. Video files are not included.
-
-Each row contains:
-- category
-- task
-- question
-- options
-- answer
-- videos
-- video_ranges
+SYNCR is a synthetic cross-video reasoning benchmark for evaluating multimodal large language models on questions that require reasoning across multiple independent videos.
+
+The benchmark contains 8,163 multiple-choice question-answer examples spanning 4 reasoning categories and 8 tasks. Each example is programmatically grounded in synthetic video data generated from CLEVRER, Kubric, and Habitat-based environments.
+
+## Dataset contents
+
+The Hugging Face dataset viewer exposes the annotation file:
+
+- `test/metadata.jsonl`
+
+Large video assets are provided in this repository as archive files and/or task-specific folders. The `videos` field in each annotation row points to the video files associated with that question.
+
+## Reasoning categories and tasks
+
+SYNCR covers four reasoning categories:
+
+1. Temporal Alignment
+   - CLEVRER temporal ordering
+   - Kubric synchronization
+
+2. Spatial Tracking
+   - Kubric spatial tracking
+   - Habitat agent/object tracking
+
+3. Comparative Reasoning
+   - CLEVRER kinematic comparison
+   - CLEVRER collision comparison
+
+4. Holistic Synthesis
+   - Habitat object counting
+   - Habitat route planning
+
+## Splits
+
+The current release provides one split:
+
+- `test`: 8,163 examples
+
+## Schema
+
+Each row contains the following fields:
+
+- `category`: the high-level reasoning category
+- `task`: the specific benchmark task
+- `question`: the natural-language multiple-choice question
+- `options`: the answer choices
+- `answer`: the correct answer
+- `videos`: the list of video paths associated with the question
+- `video_ranges`: optional temporal ranges or clip spans associated with the videos
+
+## Intended use
+
+SYNCR is intended for evaluating cross-video reasoning in multimodal models. It is designed for diagnostic evaluation rather than model training, with programmatically generated questions and answers that support controlled analysis of spatial, temporal, comparative, and holistic reasoning failures.
+
+## License
+
+This dataset release is provided under the MIT license. If the paper, code, or generated video assets use different licenses, please refer to the corresponding repository files for those terms.