Tag
This paper addresses open questions in the Gold-Angluin model of language identification in the limit, showing that computational traces using only a small alphabet and defined directly from the language enable identification in the limit, without requiring an underlying machine model.
This paper formalizes the concept of prompting complexity, which measures the shortest plausible prompt required for a fixed language model to produce a target text or behavior, drawing an analogy to resource-bounded Kolmogorov complexity.
This paper proposes a framework to distinguish between capability elicitation and creation in large language model post-training using a free-energy perspective, arguing that supervised fine-tuning and reinforcement learning often reweight existing behaviors rather than creating new ones.