Update README.md

README.md

@@ -10,7 +10,9 @@ license: cc-by-nc-sa-4.0
 ---
 ## Model Description
 
-VL-LN Bench is the first benchmark for **Interactive Instance Object Navigation (IION)**, where an embodied agent must locate a specific object instance in a realistic 3D home while engaging in **free-form natural-language dialogue**. It provides an **automated data-collection pipeline** to generate large-scale training data. Using this dataset, we train a **IION base model** with the same architecture as **InternVLA-N1**.
+VL-LN Bench is the first benchmark for **Interactive Instance Object Navigation (IION)**, where an embodied agent must locate a specific object instance in a realistic 3D home while engaging in **free-form natural-language dialogue**. It also provides an **automated data-collection pipeline** that generates large-scale training data for learning interactive navigation behaviors. Using this dataset, we train an **IION base model** that shares the same architecture as **InternVLA-N1**.
+
+The resulting model demonstrates baseline competence on IION: it can search for a specific instance in **previously unseen** environments. During exploration, the agent can either **move** by predicting a pixel-goal waypoint or **ask** a question to reduce ambiguity and improve task success and efficiency.
 
 ---
 ### 🔗 Resources