Publish ModernBERT-CNN with concise usage, training details and own evaluation metrics
90316bb verified Download evaluation/human_metrics.json from AhmedSSabir/ModernBERT-CNN-Caption-Reranker: direct link, hf CLI and curl.
- Browser
- Download file 879 Bytes
-
https://huggingface.co/AhmedSSabir/ModernBERT-CNN-Caption-Reranker/resolve/main/evaluation/human_metrics.json
- Command line
-
hf download hf://AhmedSSabir/ModernBERT-CNN-Caption-Reranker/evaluation/human_metrics.json
-
curl -L -o human_metrics.json https://huggingface.co/AhmedSSabir/ModernBERT-CNN-Caption-Reranker/resolve/main/evaluation/human_metrics.json
879 Bytes
| { | |
| "dataset": "Original Flickr8K ExpertAnnotations.txt", | |
| "pairs": 5822, | |
| "images": 1000, | |
| "checkpoint_sha256": "13502ce4a3578a94effd914525935a2978facd1c7cb243526c3161cbd12db5c2", | |
| "metrics": { | |
| "examples": 5822, | |
| "spearman": 0.2936773447142244, | |
| "kendall_tau_b": 0.21296227698549494, | |
| "kendall_tau_c": 0.2081649490635248, | |
| "pearson": 0.37538085539793303, | |
| "normalized_human_mae": 0.22123509002405464, | |
| "empty_image_contexts": 0 | |
| }, | |
| "protocol": "Original mean human ratings; three fixed image-only context policies; no tuning on human ratings", | |
| "limitations": [ | |
| "Visual extractors differ from training-data extraction; not original-paper replication", | |
| "Normalized-human MAE is descriptive; binary-trained outputs are not calibrated human ratings", | |
| "Image-context ablations share test data and are not independent validation sets" | |
| ] | |
| } | |