Assessing True Type Inference Ability of LLMs on Java Snippets

On the leakage‑free ThaliaType benchmark, LLMs saw precision drop up to 59% and recall drop up to 72% compared to StatType‑SO results. StatType‑SO has been public on GitHub since 2017. getnews.me/assessing-true-type-infe... #typeinference #leakagefree

Leakage‑Free Evaluation Improves ML Bearing Fault Diagnosis

A study submitted on 26 Sep 2025 to a journal proposes bearing-wise, leakage-free evaluation and multi-label fault detection, tested on CWRU, PU and UORED-VAFCLS datasets. getnews.me/leakage-free-evaluation-... #bearingfault #leakagefree