card: link the FGIR-Backbones collection

README.md

@@ -64,5 +64,6 @@ ckpt = torch.load(path, map_location='cpu', weights_only=False)   # dict: config
 Part of the [ERISLab FGIRFT collection](https://huggingface.co/collections/ERISLab/fgirft-full-fine-tuning-benchmark-for-fgir-6ab2e8940f3a28ec7ba0bfbf).
 For the parameter-efficient adapter works (ILA, SAW, AAA) on a frozen ViT-B/16, see
 [`ERISLab/FGIR-ViT`](https://huggingface.co/ERISLab/FGIR-ViT). As a full fine-tuning line, FGIRFT sits
-close to the [`ERISLab/FGIR-Backbones`](https://huggingface.co/ERISLab/FGIR-Backbones) benchmark,
-which studies backbones and pretraining recipes for fine-grained recognition.
+close to the [ERISLab FGIR-Backbones benchmark](https://huggingface.co/collections/ERISLab/fgir-backbones-fgvc13-cvpr-2026-6ab234cc19fcfd2de1f796a4)
+(repo [`ERISLab/FGIR-Backbones`](https://huggingface.co/ERISLab/FGIR-Backbones)), which studies
+backbones and pretraining recipes for fine-grained recognition.