A geometric relation of the error introduced by sampling a language model's output distribution to its internal state

ArXi:2605.04899v1 Announce Type: new GPT-style language models are sensitive to single-token changes at generation points where the predicted probability distribution is spread across multiple tokens. Viewing this sensitivity as a geometric property, we derive an $\mathfrak{so}(n)$-valued 1-form that depends only on the geometry of the token embeddings.