Shorten the lead

README.md

@@ -22,13 +22,9 @@ Representation Learning. EMNLP 2026, main conference.
 
 [Paper][paper] · [Project page][project] · [Code][code]
 
-PRISM separates what an action is from the viewpoint and scene it happens to be seen in.
-Language is what makes that separation trainable: descriptions recombine freely, so the
-action of one clip can be put together with the context of another into a composition that
-never occurred in the training data, and the model is asked to predict the meaning of that
-composition. Semantics that leaked between the two factors cannot survive the swap. What
-this model returns is the view-invariant side of the split, so cosine similarity measures
-what is happening rather than where the camera is.
+PRISM is a video encoder that embeds a clip by what is happening in it, not by where the
+camera is. It learns that split by recombining one clip's action with another clip's context
+in language, where such combinations come free.
 
 This checkpoint is trained on EgoExo4D, using both ego and exo views and captions generated
 by Gemini 3.