Add BioCLIP 2.5 embeddings and benchmark artifacts

bioclip2_5_benchmark/summary_20260326_185250.md

@@ -0,0 +1,35 @@
+# Full Benchmark: bioclip2_5
+
+- Timestamp: 20260326_185250
+- Samples: 139,067
+- Classes: 60
+- Reductions requested: tsne, umap, pca
+- Reductions completed: tsne, umap, pca
+- Clustering: hdbscan, dbscan, hierarchical, gmm
+
+## Skipped Clustering Runs
+
+- tsne + hierarchical (hierarchical_k60): skipped: n_samples=139067 exceeds max_samples_hierarchical=50000
+- umap + hierarchical (hierarchical_k60): skipped: n_samples=139067 exceeds max_samples_hierarchical=50000
+- pca + hierarchical (hierarchical_k60): skipped: n_samples=139067 exceeds max_samples_hierarchical=50000
+
+## Top Results by V-Measure
+
+| Rank | Reduction | Clustering | Config | V-Measure | AMI | ARI | Outlier % |
+|---:|---|---|---|---:|---:|---:|---:|
+| 1 | tsne | hdbscan | hdbscan_mcs150_ms50 | 0.8822 | 0.8814 | 0.7963 | 13.59 |
+| 2 | umap | hdbscan | hdbscan_mcs150_ms50 | 0.8590 | 0.8580 | 0.7149 | 15.28 |
+| 3 | umap | dbscan | dbscan_relaxed_ms20 | 0.7992 | 0.7971 | 0.4547 | 6.28 |
+| 4 | tsne | gmm | gmm_k60 | 0.7735 | 0.7728 | 0.5789 | 0.00 |
+| 5 | umap | gmm | gmm_k60 | 0.7727 | 0.7720 | 0.5476 | 0.00 |
+| 6 | tsne | dbscan | dbscan_relaxed_ms20 | 0.7717 | 0.7693 | 0.2577 | 2.90 |
+| 7 | pca | gmm | gmm_k60 | 0.3244 | 0.3223 | 0.0944 | 0.00 |
+| 8 | pca | hdbscan | hdbscan_mcs150_ms50 | 0.1804 | 0.1802 | 0.0130 | 1.06 |
+| 9 | pca | dbscan | dbscan_relaxed_ms20 | 0.0091 | 0.0056 | 0.0000 | 4.97 |
+
+## Artifact Paths
+
+- CSV: C:\Users\Hugo\Documents\zeroshot_clustering_animal_images\results\benchmark_bioclip2_5_full_20260326_185250\benchmark_results.csv
+- JSON: C:\Users\Hugo\Documents\zeroshot_clustering_animal_images\results\benchmark_bioclip2_5_full_20260326_185250\benchmark_results.json
+- Reductions: C:\Users\Hugo\Documents\zeroshot_clustering_animal_images\outputs\reductions
+- Clustering: C:\Users\Hugo\Documents\zeroshot_clustering_animal_images\outputs\clustering