Autoregressive next-token learning selects for world-tracking representations in LLMs if and only if the encoding preserves the training ecology's equivalence classes, as shown by decomposing cross-entropy into irreducible conditional entropy plus a Jensen-Shannon excess that vanishes exactly on the
Jack Lindsey
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
LLM chat systems show large differences in reference quantity and quality, but users rarely click or engage with them.
citing papers explorer
-
Task Ecologies and the Evolution of World-Tracking Representations in Large Language Models
Autoregressive next-token learning selects for world-tracking representations in LLMs if and only if the encoding preserves the training ecology's equivalence classes, as shown by decomposing cross-entropy into irreducible conditional entropy plus a Jensen-Shannon excess that vanishes exactly on the
-
Analyzing the Presentation, Content, and Utilization of References in LLM-powered Conversational AI Systems
LLM chat systems show large differences in reference quantity and quality, but users rarely click or engage with them.