docs(readme): describe dataset composition

README.md

@@ -16,27 +16,54 @@ tags:
 - multimodal-large-language-models
 ---
 
-# NeSy-Route
+<div align="center">
 
-This Hugging Face repository hosts the dataset release for **NeSy-Route**, a neural-symbolic benchmark for constrained route planning in remote-sensing imagery.
+<h1>NeSy-Route Dataset</h1>
 
-The dataset contains data for three benchmark tasks:
+<p><strong>A neural-symbolic benchmark for constrained route planning in remote-sensing imagery.</strong></p>
 
-- **Task 1** evaluates few-shot semantic traversability and cost-vector prediction.
-- **Task 2** evaluates zero-shot constraint-aware semantic and region reasoning.
-- **Task 3** evaluates zero-shot constrained route planning with predicted waypoint or trajectory outputs.
+<p>
+  <a href="https://mingyang1010.github.io/NeSy-Route/">
+    <img src="https://cdn.simpleicons.org/googlechrome/34A853" alt="Project Page" height="18">
+    <b>Project Page</b>
+  </a>
+  &nbsp;&nbsp;|&nbsp;&nbsp;
+  <a href="https://arxiv.org/abs/2603.16307">
+    <img src="https://cdn.simpleicons.org/arxiv/B31B1B" alt="arXiv" height="18">
+    <b>Paper</b>
+  </a>
+  &nbsp;&nbsp;|&nbsp;&nbsp;
+  <a href="https://github.com/MingYang1010/NeSy-Route">
+    <img src="https://cdn.simpleicons.org/github/181717" alt="GitHub" height="18">
+    <b>Code</b>
+  </a>
+</p>
 
-Evaluation code, prompt templates, and running instructions are available in the GitHub repository:
+</div>
 
-<https://github.com/MingYang1010/NeSy-Route>
+## Dataset Composition
+
+NeSy-Route is organized into three benchmark tasks for evaluating perception, symbolic reasoning, and constrained route planning over remote-sensing imagery.
 
-Project page:
+| Task | Setting | What It Evaluates | Split / Level | Samples |
+| --- | --- | --- | --- | ---: |
+| Task 1 | Few-shot | Semantic traversability and cost-vector prediction | Main set | 3,607 |
+| Task 2 | Zero-shot | Constraint-aware semantic and region reasoning | Level 1 | 7,659 |
+| Task 2 | Zero-shot | Constraint-aware semantic and region reasoning | Level 2 | 3,712 |
+| Task 2 | Zero-shot | Constraint-aware semantic and region reasoning | Level 3 | 1,604 |
+| Task 2 | Zero-shot | Constraint-aware semantic and region reasoning | Total | 12,975 |
+| Task 3 | Zero-shot | Constrained route planning with waypoint or trajectory outputs | Easy | 6,492 |
+| Task 3 | Zero-shot | Constrained route planning with waypoint or trajectory outputs | Medium | 2,705 |
+| Task 3 | Zero-shot | Constrained route planning with waypoint or trajectory outputs | Hard | 1,624 |
+| Task 3 | Zero-shot | Constrained route planning with waypoint or trajectory outputs | Total | 10,821 |
 
-<https://mingyang1010.github.io/NeSy-Route/>
+Task 1 and Task 2 provide semantic constraint annotations for vector-based reasoning. Task 3 provides route-planning samples together with traversability maps, cost maps, ground-truth paths, and difficulty-level subsets.
 
-Paper:
+## Evaluation Code
 
-<https://arxiv.org/abs/2603.16307>
+Evaluation scripts, prompt templates, and running instructions are available in the GitHub repository:
+
+<https://github.com/MingYang1010/NeSy-Route>
 
 ## License