https://tanzglobal.com/computer-vision-test-set/2026-10-12T14:23:19+00:00https://tanzglobal.com/check-ai-summaries/2026-10-12T14:23:21+00:00https://tanzglobal.com/ai-context-window-memory/2026-10-12T19:45:47+00:00https://tanzglobal.com/ai-abstention-uncertainty-design/2026-10-13T08:38:31+00:00https://tanzglobal.com/machine-learning-data-leakage/2026-10-13T19:59:14+00:00https://tanzglobal.com/an-ai-confidence-score-needs-a-reality-check/2026-10-14T12:01:44+00:00https://tanzglobal.com/ask-which-error-matters-before-comparing-ai-accuracy/2026-10-14T20:37:43+00:00https://tanzglobal.com/ocr-can-read-every-number-and-still-put-it-in-the-wrong-row/2026-10-15T19:13:13+00:00https://tanzglobal.com/a-benchmark-average-can-hide-the-language-your-users-need/2026-10-15T20:33:45+00:00https://tanzglobal.com/a-public-ai-dataset-still-needs-a-usage-check/2026-10-15T22:27:11+00:00https://tanzglobal.com/data-drift-can-change-an-ai-system-without-a-model-update/2026-10-16T19:40:57+00:00https://tanzglobal.com/a-model-card-helps-you-ask-what-an-ai-system-was-tested-for/2026-10-17T00:26:58+00:00https://tanzglobal.com/synthetic-data-does-not-automatically-remove-privacy-risk/2026-10-17T20:21:17+00:00https://tanzglobal.com/a-faster-ai-model-is-useful-only-if-you-measure-the-whole-task/2026-10-17T23:10:09+00:00https://tanzglobal.com/an-ai-search-result-needs-a-retrieval-date-as-well-as-a-source/2026-10-18T23:02:04+00:00https://tanzglobal.com/semantic-similarity-does-not-mean-two-sentences-agree/2026-10-19T01:27:43+00:00https://tanzglobal.com/a-language-model-s-temperature-is-not-a-truthfulness-setting/2026-10-19T21:33:18+00:00https://tanzglobal.com/a-quantized-ai-model-needs-its-own-quality-check/2026-10-20T00:00:08+00:00https://tanzglobal.com/a-token-count-is-not-a-word-count/2026-10-20T22:30:39+00:00https://tanzglobal.com/training-and-inference-put-an-ai-model-to-different-kinds-of-work/2026-10-21T02:11:22+00:00https://tanzglobal.com/an-embedding-s-dimensions-are-coordinates-not-named-human-traits/2026-10-21T21:06:22+00:00https://tanzglobal.com/where-you-split-a-document-can-change-an-ai-search-answer/2026-10-22T09:25:38+00:00https://tanzglobal.com/changing-a-classification-threshold-changes-which-ai-mistakes-you-see/2026-10-22T19:05:30+00:00https://tanzglobal.com/confusion-matrix-classification/2026-10-23T00:22:15+00:00https://tanzglobal.com/context-window-output-budget/2026-10-23T21:44:25+00:00https://tanzglobal.com/ai-random-seed-reproducibility/2026-10-24T06:29:29+00:00https://tanzglobal.com/macro-micro-classification/2026-10-24T18:52:59+00:00https://tanzglobal.com/stratified-data-split/2026-10-25T07:10:25+00:00https://tanzglobal.com/precision-recall-positive-label/2026-10-25T23:46:10+00:00https://tanzglobal.com/grouped-ai-data-split/2026-10-26T00:54:41+00:00https://tanzglobal.com/mae-rmse-large-errors/2026-10-26T22:31:59+00:00https://tanzglobal.com/time-series-train-test-split/2026-10-27T12:15:17+00:00https://tanzglobal.com/an-unseen-category-needs-a-plan-before-one-hot-encoding-goes-live/2026-10-27T21:35:28+00:00https://tanzglobal.com/cosine-similarity-compares-direction-while-a-dot-product-also-keeps-magnitude/2026-10-28T11:57:42+00:00https://tanzglobal.com/a-decision-tree-s-maximum-depth-changes-how-finely-it-divides-the-data/2026-10-28T18:38:24+00:00https://tanzglobal.com/permutation-importance-can-lose-clarity-when-features-overlap/2026-10-29T05:32:51+00:00https://tanzglobal.com/class-weights-change-the-learning-penalty-not-the-population-you-measure/2026-10-29T23:13:09+00:00https://tanzglobal.com/learning-curves-help-you-question-whether-the-next-experiment-needs-more-data/2026-10-30T09:56:54+00:00https://tanzglobal.com/choosing-when-to-stop-training-without-using-the-final-test-set/2026-10-31T01:25:32+00:00https://tanzglobal.com/multilabel-classification-lets-one-example-belong-to-several-classes/2026-10-31T11:39:42+00:00https://tanzglobal.com/feature-scaling-changes-what-distance-means-to-k-means-clustering/2026-10-31T21:41:06+00:00https://tanzglobal.com/dbscan-can-identify-noise-without-choosing-a-cluster-count-first/2026-11-01T01:14:27+00:00https://tanzglobal.com/pca-finds-directions-of-variation-not-the-causes-behind-your-data/2026-11-02T11:19:40+00:00https://tanzglobal.com/top-p-sampling-changes-which-tokens-an-ai-model-may-choose-next/2026-11-02T14:52:42+00:00https://tanzglobal.com/a-language-model-s-kv-cache-saves-repeated-work-during-generation/2026-11-02T22:11:58+00:00https://tanzglobal.com/mixture-of-experts-models-activate-a-subset-of-their-capacity-for-each-token/2026-11-03T05:58:49+00:00