A Theory of LLM Information Susceptibility
This provides predictive constraints for designing AI systems, clarifying when LLM intervention is beneficial, which is incremental but addresses a fundamental limit in deploying LLMs as optimization modules.
The paper tackles the problem of understanding when large language models (LLMs) improve performance in agentic systems, showing that fixed LLMs do not increase performance susceptibility with sufficient computational resources, but nested co-scaling architectures can exceed this bound and open new response channels.
Large language models (LLMs) are increasingly deployed as optimization modules in agentic systems, yet the fundamental limits of such LLM-mediated improvement remain poorly understood. Here we propose a theory of LLM information susceptibility, centred on the hypothesis that when computational resources are sufficiently large, the intervention of a fixed LLM does not increase the performance susceptibility of a strategy set with respect to budget. We develop a multi-variable utility-function framework that generalizes this hypothesis to architectures with multiple co-varying budget channels, and discuss the conditions under which co-scaling can exceed the susceptibility bound. We validate the theory empirically across structurally diverse domains and model scales spanning an order of magnitude, and show that nested, co-scaling architectures open response channels unavailable to fixed configurations. These results clarify when LLM intervention helps and when it does not, demonstrating that tools from statistical physics can provide predictive constraints for the design of AI systems. If the susceptibility hypothesis holds generally, the theory suggests that nested architectures may be a necessary structural condition for open-ended agentic self-improvement.