Org card: update Fusion Perception to the shipped retrieval model

README.md

@@ -35,7 +35,15 @@ semantic search. The senses above attach to these and never modify them.
 
 ## Fusion Perception, the perception layer
 
-Dense scene understanding and geometric place recognition on a frozen vision backbone, projected into the Fusion Embedding space. Code: https://github.com/Eximius-Labs/fusion-perception.
+Place recognition: has this robot been here before. A 3.1M-parameter head on a frozen
+DINOv2 ViT-L/14 backbone turns an image into one 512-d descriptor, for re-localization
+against a large index. Searching 1,001,001 distractor photographs, it reaches 92.61 mAP
+Medium and 86.03 Hard on RParis+1M with AMES reranking, above the best published numbers
+we are aware of; on ROxford+1M it is about three points behind that frontier.
+
+- [fusion-perception-1](https://huggingface.co/EximiusLabs/fusion-perception-1-preview) — the retrieval head, with both the standard and decontaminated training protocols.
+
+Code: https://github.com/Eximius-Labs/fusion-perception
 
 ## Engram, the memory layer