Attribution, Temporality, and Modality as Compact Causal Indices in Large Language Models

This repository contains a hypothesis paper proposing a unified computational view of attribution (who), temporality (past/future), and modality (factual/counterfactual) in large language models. We argue that these distinctions are not semantic content but compact causal indexing variables that regulate how event descriptions are applied to internal world-state updates during generation. The work is released under CC BY 4.0 to encourage reuse, critique, and extension of the proposed hypothesis.