RiverRider/srt-cxr14-frozen-probe

Frozen general-purpose features beat a fine-tuned baseline on ChestX-ray14 A linear probe on frozen google/gemma-4-31B-it hidden states, on all 112,120 images of NIH ChestX-ray14, using the official test_list.txt. No fine-tuning, no radiology pretraining, no augmentation. One Linear(d, 14) under BCE, which is fourteen logistic regressions. ChestX-ray14, official split mean AUROC method Wang et al. 2017 (dataset authors) 0.7451 ResNet-50, fine-tuned end to end this… See the full description on the dataset page: https://huggingface.co/datasets/RiverRider/srt-cxr14-frozen-probe.

<iframe src="https://silytics.si/embed/item/eca84201-c162-41f6-a8c0-29011991446f" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.