LATIDIA · Investigación

¿Cuántos pensamientos puede contener un vector? La capacidad de razonamiento por superposición

arXiv: 2609.13747v1Tipo de anuncio: nuevo Resumen: Los modelos de lenguaje grandes resuelven problemas difíciles a través de cálculos intermedios a través del razonamiento de varios pasos. La cadena de pensamiento tradicional codifica estos cálculos como tokens.

La noticia

arXiv:2609.13747v1 Announce Type: new Abstract: Large language models solve hard problems through intermediate computations across multi-step reasoning. Traditional chain-of-thought encodes these computations as tokens. Recent continuous and recurrent methods instead move partial computations into fixed-dimensional latent states, where a single thought can superpose multiple alternatives. This raises a fundamental design question:what should continuous thoughts preserve as reasoning proceeds? An intuitive approach discards past computations and keeps only the current reasoning frontier. Storing more items seems to dilute states and waste limited representational capacity. We show this intuition can be incorrect. Under identical downstream computations, cumulative superposition retaining full reasoning history can require lower representational dimensions than frontier-only superposition holding only