Add paper and code links, update citation

README.md

@@ -1,11 +1,13 @@
 ---
-pretty_name: FlowBench
-license: other # TODO: set the release licence; source frames keep their original licences (see "Licences")
+language:
+- en
+license: other
+size_categories:
+- 100K<n<1M
 task_categories:
 - visual-question-answering
 - image-text-to-text
-language:
-- en
+pretty_name: FlowBench
 tags:
 - spatial-reasoning
 - counterfactual-reasoning
@@ -15,8 +17,6 @@ tags:
 - set-of-marks
 - multiple-choice
 - ai2thor
-size_categories:
-- 100K<n<1M
 configs:
 - config_name: sim_colour_som
   default: true
@@ -71,6 +71,8 @@ measured 3D geometry for FlowBench-Real. No model chooses the answers. For FlowB
 ground truth behind every question (optical flow, depth, instance segmentation, moved-object masks, camera and
 move metadata) is released as a separate config, `sim_ground_truth`.
 
+**Paper:** [FlowBench: Counterfactual Imagination Training for Spatial Intelligence](https://huggingface.co/papers/2611.03937) · **Code:** [github.com/liuyifan22/FlowBench](https://github.com/liuyifan22/FlowBench)
+
 | subset | scenes | split | questions | image pairs | image styles |
 |---|---|---|---|---|---|
 | **FlowBench-Sim** | AI2-THOR indoor houses | `train` (ProcTHOR-10k) | 419,612 | 117,604 scene samples | `natural`, `colour_som`, `red_som` |
@@ -146,9 +148,13 @@ Options are stored in separate columns, and unused letters are `null`. A simple
 
 ```python
 letters = [l for l in "ABCDEF" if row[f"choice_{l}"] is not None]
-options = "\n".join(f"{l}. {row[f'choice_{l}']}" for l in letters)
-counterfactual_prompt = f"{row['motion_command']}\n{row['question']}\n{options}"  # with image_before
-static_prompt = f"{row['static_question']}\n{options}"                           # with image_after
+options = "
+".join(f"{l}. {row[f'choice_{l}']}" for l in letters)
+counterfactual_prompt = f"{row['motion_command']}
+{row['question']}
+{options}"  # with image_before
+static_prompt = f"{row['static_question']}
+{options}"                           # with image_after
 ```
 
 FlowBench-Real rows also include the exact text used in our evaluations: `prompt` for the counterfactual
@@ -411,10 +417,11 @@ scale reference measured the same way ("For scale, X is about N cm across at its
 ## Citation
 
 ```bibtex
-@misc{flowbench,
-  title  = {FlowBench},
-  author = {TODO},
-  year   = {2026},
-  note   = {TODO: add paper reference}
+@misc{liu2026flowbench,
+  title        = {FlowBench: Counterfactual Imagination Training for Spatial Intelligence},
+  author       = {Liu, Yifan and Zhang, Baihan and Aw, Khai Loong and Kotar, Klemen and Yamins, Daniel L. K.},
+  year         = {2026},
+  howpublished = {\url{https://github.com/liuyifan22/FlowBench}},
+  note         = {Yifan Liu, Baihan Zhang, Khai Loong Aw, and Klemen Kotar contributed equally. Paper: \url{https://huggingface.co/papers/2611.03937}}
 }
-```
+```