Update README

README.md

@@ -13,3 +13,40 @@ configs:
 ---
 
 # SpatialTunnel
+
+SpatialTunnel is a Blender-rendered diagnostic dataset for studying how vision-language models represent spatial relations internally. It accompanies the contrastive-probing framework introduced in **Why Far Looks Up: Probing Spatial Representation in Vision-Language Models**.
+
+Links:
+
+- Contrastive-probing code and methodology: https://github.com/cheolhong0916/contrastive-probing
+- Project page: https://cheolhong0916.github.io/whyfarlooksup.github.io/
+- SpatialTunnel generation code: https://github.com/cube-c/spatialtunnel-dataset-gen
+
+The benchmark uses simple rendered 3D scenes to isolate spatial-shortcut biases.
+
+## Dataset Configs
+
+| Config | File | Rows | Description |
+|---|---:|---:|---|
+| `size_variation` | `size_variation.tsv` | 1,100 | Apparent-size questions under controlled object-size variation. |
+| `phase_variation` | `phase_variation.tsv` | 3,072 | Distance questions with controlled angular-position variation. |
+| `contrastive_probing` | `contrastive_probing.tsv` | 1,200 | Balanced spatial-relation questions for contrastive probing. |
+
+## Format
+
+All configs are tab-separated files with the same set of columns:
+
+| Column | Description |
+|---|---|
+| `index` | Row index within the selected config. |
+| `image` | Base64-encoded PNG image. |
+| `question` | Spatial question to ask the model. |
+| `A` | First answer choice. |
+| `B` | Second answer choice. |
+| `category` | Gold label for the row. |
+
+## Generation
+
+SpatialTunnel scenes are generated in Blender using the separate dataset-generation repository:
+
+https://github.com/cube-c/spatialtunnel-dataset-gen