Emergence Invariance: Can Scaling Compensate for LLM Incompleteness?
A new paper on arXiv (2608.01548) proposes the Symbolization–Substructure Thesis and introduces the concept of emergence invariance to question whether scaling large language models can compensate for all missing distinctions relative to human cognition. The authors argue that language is a formalized subset of thought, and while LLMs exhibit compensatory emergence—sparse architectural primitives enabling in-context learning, multi-step reasoning, tool use, and chain of thought—they still suffer from substantive, substrate, and high-level incompletenesses. The paper formalizes this with the equation R_s* = R_phi* + C_s, suggesting that scale can reduce the complexity gap but not eliminate it. The study is theoretical, drawing on philosophy and AI research, and was announced as a new submission on arXiv.
Key facts
- Paper arXiv:2608.01548, announced as new on arXiv.
- Introduces the Symbolization–Substructure Thesis.
- Defines emergence invariance as a concept.
- Equation: R_s* = R_phi* + C_s.
- Claims LLMs exhibit compensatory emergence.
- Identifies three types of incompleteness: substantive, substrate, and high-level.
- Questions whether emergence can compensate for every missing distinction.
- Published on arXiv, a preprint server.
Entities
Institutions
- arXiv