Tighten RF100-VL cross-domain benchmark section

README.md

@@ -123,7 +123,7 @@ The detection head wins 17 of 20 domains, loses 3, with mean AR@100 +3.8 over th
 
 ### Cross-domain detection benchmark (RF100-VL subset)
 
-The same 20-domain RF100-VL class-agnostic AR@100 protocol applied to standard 2026 detector families: YOLO26 (n/s/m/l/x) from Ultralytics and RF-DETR (Nano/Small/Medium/Large) from Roboflow. All models trained on COCO; none saw RF100-VL. Each model evaluated at its native input resolution (YOLO26 variants at 768, RF-DETR at 384/512/576/704 for Nano/Small/Medium/Large, Argus at 768), max 100 detections per image, class labels ignored. Columns after Argus are sorted by mean AR@100 descending.
+YOLO26 (n/s/m/l/x) and RF-DETR (Nano/Small/Medium/Large) on the same 20 RF100-VL domains, same class-agnostic AR@100 protocol. All trained on COCO only.
 
 | domain | Argus+FCOS | Argus+(current) | RF-DETR-L | RF-DETR-M | RF-DETR-S | YOLO26l | RF-DETR-N | YOLO26x | YOLO26m | YOLO26s | YOLO26n |
 |---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
@@ -150,8 +150,6 @@ The same 20-domain RF100-VL class-agnostic AR@100 protocol applied to standard 2
 | **mean AR@100**           | **25.1** | **28.9** | **34.4** | **33.3** | **32.3** | **31.2** | **30.8** | **30.5** | **29.8** | **29.2** | **28.8** |
 | **total params**          | 102.1M | 89.0M | 33.9M | 33.7M | 32.1M | 26.3M | 30.5M | 59.0M | 21.9M | 10.0M | 2.6M |
 
-The Argus total-params entries (102.1M and 89.0M) include the 86M EUPE-ViT-B backbone shared across five task heads. YOLO26 and RF-DETR entries are standalone detection models with their own backbones.
-
 ### Cross-Dataset Segmentation Transfer
 
 A separate BN+1×1 linear probe with the same training recipe as the ADE20K head, on the frozen backbone. The backbone was never exposed to driving scenes during EUPE distillation or Argus head training.