Vitality Watch

⚠ Evidence note: An aging clock is a statistical model, not a measurement of aging itself; predictive validity varies widely between clocks and populations.

Mentioned 1× · first seen 28-sep-2026 · last seen 28-sep-2026

LongevityBench and Longevity-LLMs (Aging-Biology AI Benchmark)

Current understanding

LongevityBench is a first-of-its-kind open multi-domain benchmark for testing whether frontier large language models can interpret heterogeneous aging biodata (including omics age prediction, which the benchmark reports as the hardest task). Alongside it, compact fine-tuned Longevity-LLMs (0.6B–9B parameters) reportedly match or exceed much larger general-purpose systems and ship with an agentic research interface [expert-opinion]. As a measurement-layer tooling contribution it addresses how AI over aging biomarkers is evaluated, not whether any given clock predicts outcomes.

Evidence log

Open questions

Related

Contributing findings

pubmed other
An open benchmark and language models for AI in aging biology.
01-sep-2026 novelty 4 per-area 4 introduces

build build 880 · 2bcef25 · 2026-11-18