# Autoregressive (next-token) modelling

Source: https://onco.cc/terms/autoregressive-modelling/  
OnCo record `autoregressive-modelling` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

An autoregressive model predicts the next element of a sequence from the ones before it; trained on DNA it learns to continue a genome, which is how Evo 2 was built.

## Summary

An autoregressive model expresses each output as a function of its own previous values (Wikipedia); large language models are trained this way on text to generate the next token (Wikipedia on LLMs). Genomic models such as Evo 2 and the Nucleotide Transformer apply it to nucleotide sequence at up to a million bases of context, and the likelihood the model assigns to a variant sequence is used as a zero-shot pathogenicity score. Generation and scoring are different uses, and success at one does not imply the other.

## Fields

- Kind: Term
- Last checked: 2026-09-24
- Also known as: autoregressive modelling; autoregressive model; next-token prediction; next token prediction; causal language modelling; generative pretraining
- Tags: cansim-terms

## Notes

- Listed in the CanSim terms map 1.0.0 (docs/onco/terms.json, generated 2026-09-24), CC BY 4.0, attribution: CanSim project, an open, public-data-first cancer foundation-model programme; CanSim page path /terms/autoregressive-modelling.

## Sources

- Wikipedia: https://en.wikipedia.org/wiki/Autoregressive_model
- Wikipedia: large language model: https://en.wikipedia.org/wiki/Large_language_model
- Wikipedia: https://en.wikipedia.org/wiki/Autoregressive_model

## Connected records

- terms: [Cancer AI vocabulary (CanSim terms map)](https://onco.cc/terms/cancer-ai-vocabulary/), [Genomic and protein language models: Evo 2, Enformer, ESM](https://onco.cc/terms/genomic-and-protein-language-models/), [Transformer and attention](https://onco.cc/terms/transformer-architecture/), [Variant effect prediction](https://onco.cc/terms/variant-effect-prediction/), [Zero-shot prediction](https://onco.cc/terms/zero-shot/)
- technologies: [Evo 2 (Arc Institute, NVIDIA)](https://onco.cc/technologies/evo2/), [Nucleotide Transformer (InstaDeep)](https://onco.cc/technologies/nucleotide-transformer/)

---
JSON: https://onco.cc/api/v1/entities/autoregressive-modelling.json