https://yuriihavrylko.github.io/Yurii HavrylkoNotes on machine learning, AI, and LLM engineering — benchmarks, performance, and production systems. 2026-07-09T12:00:30+00:00 Yurii Havrylko https://yuriihavrylko.github.io/ Jekyll © 2026 Yurii Havrylko /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png Evaluating Zero-Shot NER: What It Takes to Not Fool Yourself2026-07-05T16:41:00+00:00 2026-07-05T16:41:00+00:00 https://yuriihavrylko.github.io/posts/nerbench-zero-shot-ner-leaderboard/ Yurii Havrylko GLiNER, GLiNER2, and prompted LLMs on 46 NER datasets — and the three ways a zero-shot NER leaderboard quietly misleads you (unfair aggregation, no uncertainty, inflated absolutes), plus how to fix each. Benchmarking NER Inference Optimizations in Presidio: What Actually Made It Faster2026-06-21T12:32:00+00:00 2026-06-21T12:32:00+00:00 https://yuriihavrylko.github.io/posts/presidio-ner-speed-benchmark/ Yurii Havrylko ONNX Runtime, batch inference, and tokenizer-based chunking for Presidio's NER recognizers, benchmarked across five hardware platforms. The winning configuration flips with hardware class.