A single observed event sequence suffices to recover a summary causal graph if a pretrained autoregressive model is used as a conditional-mutual-information estimator, with causal error bounded by the model's cross-entropy gap.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Your Autoregressive Model Already Reveals the Causal Graph
A single observed event sequence suffices to recover a summary causal graph if a pretrained autoregressive model is used as a conditional-mutual-information estimator, with causal error bounded by the model's cross-entropy gap.