Update README.md

README.md

@@ -158,7 +158,7 @@ size_categories:
 - 🤖 **Embodiments:** Four embodiment types capturing distinct physical and spatial constraints (human, legged robot, wheeled robot, or bicycle).
 - 📏 **Scale:** 1,002 diverse real-world scenarios and over 3,000 expert-annotated traces.
 - ⚖️ **Splits:**
-  - Validation split (~50%) for experimenting and model fine-tuning.
+  - Validation split (~50%) for experimentation and model fine-tuning.
   - Test split (~50%) with hidden ground-truths for public leaderboard evaluation.
 - 🔎 **Annotation Quality:** All images and traces manually collected and labeled by human experts.
 - 🏅 **Evaluation Metric:** Semantic-aware Trace Score, combining Dynamic Time Warping distance, goal endpoint error, and embodiment-conditioned semantic penalties.