docs(README): DAVIS S/M/L table, Publisher column, updated metrics

Replace Params column with Publisher (Meta/Google/Motif); move DAVIS to front and split into S/M/L per DINOv3 Table 5; fill baseline numbers from Table 5; Motif DAVIS 74.0/80.5/83.5, ImageNet 87.2, ADE20K 52.0.

README.md

@@ -90,16 +90,21 @@ Compared against the strongest publicly reported self-supervised / vision backbo
 better for every column **except KITTI depth MSE** (lower is better). Best value per column in
 **bold**.
 
-| Model | Params | ImageNet-1K<br>lin. probe ↑ | ADE20K<br>mIoU ↑ | DAVIS<br>J&F ↑ | K400 ↑ | KITTI<br>depth MSE ↓ |
-|---|---|---|---|---|---|---|
-| **Motif Vision Encoder** | **7B** | 87.2 | 52.0 | **71.7** | *in progress* | *in progress* |
-| DINOv3 | 7B | **88.2** | **55.9** | 71.1 | **87.8** | **2.3** |
-| DINOv2 | 1.1B | 86.5 | 49.0 | 63.9 | 84.4 | – |
-| V-JEPA 2.1 | 2B | 85.5 | 47.9 | 69.0 | 87.7 | 3.x |
-| SigLIP2 | 2B | 84.5 | 45.4 | 56.1 | 86.9 | – |
-
-- **DAVIS video segmentation (J&F 71.7)** — best in the table, surpassing DINOv3 7B (71.1),
-  reflecting the encoder's dense, temporally-coherent patch features on video.
+DAVIS S/M/L follow the DINOv3 protocol (J&F-mean at video short side 420/480, 840/960,
+1260/1440 px). V-JEPA 2.1 is not part of the DINOv3 Table 5 tracking benchmark, so only its
+single-resolution (S) figure is available.
+
+| Model | Publisher | DAVIS S<br>J&F ↑ | DAVIS M<br>J&F ↑ | DAVIS L<br>J&F ↑ | ImageNet-1K<br>lin. probe ↑ | ADE20K<br>mIoU ↑ | K400 ↑ | KITTI<br>depth MSE ↓ |
+|---|---|---|---|---|---|---|---|---|
+| **Motif Vision Encoder** | Motif | **74.0** | **80.5** | **83.5** | 87.2 | 52.0 | *in progress* | *in progress* |
+| DINOv3 | Meta | 71.1 | 79.7 | 83.3 | **88.2** | **55.9** | **87.8** | **2.3** |
+| DINOv2 | Meta | 63.9 | 73.6 | 76.6 | 86.5 | 49.0 | 84.4 | – |
+| V-JEPA 2.1 | Meta | 69.0 | – | – | 85.5 | 47.9 | 87.7 | 3.x |
+| SigLIP2 | Google | 56.1 | 62.3 | 62.9 | 84.5 | 45.4 | 86.9 | – |
+
+- **DAVIS video segmentation (J&F)** — leads at every resolution: **74.0 (S) / 80.5 (M) /
+  83.5 (L)**, surpassing DINOv3 7B (71.1 / 79.7 / 83.3) across the board, reflecting the
+  encoder's dense, temporally-coherent patch features on video.
 - **ImageNet-1K linear probe (87.2)** and **ADE20K semantic segmentation (52.0 mIoU)** are
   second only to DINOv3 7B while ahead of DINOv2, V-JEPA 2.1, and SigLIP2.
 - **Data efficiency** — these results come from **~0.47B training samples** (448.6M images +