Add robotics task category and paper/GitHub links to dataset card

README.md

@@ -1,8 +1,10 @@
 ---
-pretty_name: BEHAVIOR ESI-Bench
 license: cc-by-4.0
 size_categories:
 - 1K<n<10K
+pretty_name: BEHAVIOR ESI-Bench
+task_categories:
+- robotics
 tags:
 - json
 - text
@@ -19,20 +21,16 @@ configs:
 
 # BEHAVIOR ESI-Bench
 
-BEHAVIOR ESI-Bench is a question dataset for evaluating embodied spatial intelligence across indoor scenes, object arrangements, physical reasoning, temporal understanding, and active exploration tasks.
-
-Each row is one question instance. The table is intentionally flattened for Hugging Face Dataset Viewer and Croissant compatibility.
+[**Project Page**](https://esi-bench.github.io/) | [**Paper**](https://huggingface.co/papers/2605.18746) | [**GitHub**](https://github.com/ESI-Bench/ESI-Bench)
 
-## Dataset Metadata
+BEHAVIOR ESI-Bench is a comprehensive benchmark for evaluating embodied spatial intelligence across indoor scenes, object arrangements, physical reasoning, temporal understanding, and active exploration tasks. Built on OmniGibson and grounded in Spelke's core knowledge systems, it spans 10 task categories and 29 subcategories. 
 
-- Creator: [ESI-Bench](https://huggingface.co/ESI-Bench)
-- Dataset URL: <https://huggingface.co/datasets/ESI-Bench/esi-bench>
-- Version: `1.0.0`
-- Published: `2026-05-05`
-- License: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)
+Unlike passive sensing benchmarks, ESI-Bench requires agents to close the perception-action loop: deciding what abilities to deploy—perception, locomotion, and manipulation—and how to sequence them to actively accumulate task-relevant evidence.
 
 ## Schema
 
+Each row is one question instance. The table is intentionally flattened for Hugging Face Dataset Viewer and Croissant compatibility.
+
 ```text
 id
 big_task
@@ -69,18 +67,43 @@ Temporal Understanding
 
 The `small_task` column stores the corresponding subtask, and `runner_task` stores the internal task module name used by the original BEHAVIOR active-exploration code.
 
+## Sample Usage
+
+To run the active exploration module using a task from the dataset, you can use the provided runner script in the [official repository](https://github.com/ESI-Bench/ESI-Bench):
+
+```bash
+# Example: Running the Explorer for a counting task using the Gemini provider
+python src/main.py \
+  --task counting \
+  --metadata "dataset/json_clean/Enumerative Perception/Spatial Segmentation/Merom_0_int/living_room_0/q_000.json" \
+  --provider gemini \
+  --model gemini-3.1-pro-preview \
+  --max-steps 30 \
+  --min-steps 1 \
+  --threshold 0.9 \
+  --results-root outputs/results \
+  --step-image-root outputs/steps \
+  --overwrite
+```
+
 ## Citation
 
-Use the following citation string when referencing this dataset:
+If you find ESI-Bench useful in your research, please cite:
 
-```text
-ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop. ESI-Bench dataset, 2026. https://huggingface.co/datasets/ESI-Bench/esi-bench
+```bibtex
+@inproceedings{hong2026esibench,
+  title     = {{ESI-Bench}: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop},
+  author    = {Hong, Yining and Liu, Jiageng and Yin, Han and Li, Manling and Guibas, Leonidas and Li, Fei-Fei and Wu, Jiajun and Choi, Yejin},
+  year      = {2026}
+}
 ```
 
+This benchmark is built on [BEHAVIOR-1K](https://behavior.stanford.edu/behavior-1k/) and [OmniGibson](https://behavior.stanford.edu/omnigibson/). Please consider citing those works as well if you use this dataset.
+
 ## Croissant
 
 Hugging Face automatically generates Croissant metadata from the Dataset Viewer once this dataset is processed. The Croissant JSON-LD endpoint is:
 
 ```text
 https://huggingface.co/api/datasets/ESI-Bench/esi-bench/croissant
-```
+```