Update README.md

README.md

@@ -14,7 +14,7 @@ The benchmark covers three core tasks:
 - **Temporal Audio Grounding (TAG)**: localize relevant time spans for a given query
 - **Targeted Audio Captioning (TAC)**: produce descriptions for specific temporal segments
 
-LAT-Bench contains approximately **40 hours of audio**, including:
+LAT-Bench contains approximately **40 hours of long-form audio**, including:
 - **25 hours in Chinese**
 - **15 hours in English**