A translator who gets distracted by every rustle in the room turns a one-hour meeting into two. That is how age-predicting tools used to work: each one understood only a single "dialect" of tests. One read chemical marks on DNA (they change with age), another read proteins in the blood, a third read doctors' notes. To add a new test, you had to build a new model.
Scientists have built a single neural network that learned all the dialects at once. It is like a translator who understands English, Chinese, and sign language. They called the network Longevity-LLM. It was shown 766,640 examples where the tests were written as ordinary text, and it learned to predict age. The prediction error for DNA methylation was 4.34 years, which is more accurate than the old tools.
Now this kind of translator is needed not just for accuracy. It brings all the tests together and speaks a language researchers can understand. This could help in the search for anti-aging drugs, like a single assistant that reads any data and explains it. The next version will not only predict age but also explain why it differs from the age in your passport.
Before training, the base model Qwen3-14B could not produce a single valid prediction on any task, because it did not know how to format its answers. After fine-tuning, it outperformed commercial systems with hundreds of billions of parameters.
The idea of a single AI that reads biological data and explains aging echoes the medical AI from the series Altered Carbon, where consciousness is transferred between bodies and biological age is controlled by technology.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.