Markov Chain Estimation with In-Context Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lepage, Simon, Mary, Jeremie, Picard, David
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912522076946432
author Lepage, Simon
Mary, Jeremie
Picard, David
author_facet Lepage, Simon
Mary, Jeremie
Picard, David
contents We investigate the capacity of transformers to learn algorithms involving their context while solely being trained using next token prediction. We set up Markov chains with random transition matrices and we train transformers to predict the next token. Matrices used during training and test are different and we show that there is a threshold in transformer size and in training set size above which the model is able to learn to estimate the transition probabilities from its context instead of memorizing the training patterns. Additionally, we show that more involved encoding of the states enables more robust prediction for Markov chains with structures different than those seen during training.
format Preprint
id arxiv_https___arxiv_org_abs_2508_03934
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Markov Chain Estimation with In-Context Learning
Lepage, Simon
Mary, Jeremie
Picard, David
Machine Learning
68T07
We investigate the capacity of transformers to learn algorithms involving their context while solely being trained using next token prediction. We set up Markov chains with random transition matrices and we train transformers to predict the next token. Matrices used during training and test are different and we show that there is a threshold in transformer size and in training set size above which the model is able to learn to estimate the transition probabilities from its context instead of memorizing the training patterns. Additionally, we show that more involved encoding of the states enables more robust prediction for Markov chains with structures different than those seen during training.
title Markov Chain Estimation with In-Context Learning
topic Machine Learning
68T07
url https://arxiv.org/abs/2508.03934