Sex-specific AI aging clocks outperform pooled models and surface signals they miss
bravo_abad · x · 2026-09-17
A look at biological aging clocks: because they learn a normative reference of 'normal' aging and measure deviation from it, how you pool data is a modeling choice. Training female and male clocks separately across organs and molecular measures generalized better than pooled models even with matched sample sizes and age distributions—and mattered beyond prediction: a genetic analysis linking sleep disorders to immune aging was significant in females but not in the sex-pooled model. Lesson for AI-for-Science: the reference population becomes part of the model.
More from Research
- Non-transformer deep learning work is being swept under the rug, researcher laments — cephaloform · 2026-09-17
- BuildingBench tests coding agents on 3D building generation, with up to 80% cost gaps — ZhitingHu · 2026-09-17
- Ask GPT-5.2 and Claude Opus 4.6 to 'Be the Null' and They Output Zero Bytes, 30/30 — rayanpal_ · 2026-09-17
- Fast classifier Jev vs LLMs: parallel computation trades generation for speed — FrankFelixAI · 2026-09-17
- RLVR-trained small model fixes 83.7% of LaTeX errors, 6x faster at 1/40 the cost — simonguozirui · 2026-09-17
- WIRED Reporter Vibe-Coded a Story Idea Generator From an Open-Source Fruit Fly Brain — nordicinst · 2026-09-17