> ## Documentation Index > Fetch the complete documentation index at: https://proto.evodesign.org/docs/llms.txt > Use this file to discover all available pages before exploring further. # Causal Models > Autoregressive models for sequence generation and scoring Autoregressive language models for biological sequences. Trained to predict the next residue or nucleotide from the sequence so far, they generate new sequences one token at a time from an optional prompt and score how likely an existing sequence is under the model. That likelihood serves as a zero-shot proxy for how natural a sequence looks, across both DNA and protein alphabets. * **Input:** a sequence to score, or a prompt to extend (DNA or protein). * **Output:** newly generated sequences, or per-sequence likelihood and perplexity scores.

Arc Institute

Salesforce Research

Profluent