Probabilistic Modeling of Latent Agentic Substructures in Deep Neural Networks
This work provides a foundational mathematical framework for understanding subagent coalescence in AI systems, with implications for alignment in agentic AI, though it appears incremental in building on existing probabilistic modeling concepts.
The authors tackled the problem of modeling intelligent agency in neural networks by developing a probabilistic theory that defines agents as outcome distributions with epistemic utility, and they proved that strict unanimity is possible with three or more outcomes but not under linear pooling or binary outcomes. They formalized an agentic alignment phenomenon in LLMs, showing that eliciting a benevolent persona induces an antagonistic counterpart, and a manifest-then-suppress strategy yields strictly larger misalignment reduction than pure reinforcement alone.
We develop a theory of intelligent agency grounded in probabilistic modeling for neural models. Agents are represented as outcome distributions with epistemic utility given by log score, and compositions are defined through weighted logarithmic pooling that strictly improves every member's welfare. We prove that strict unanimity is impossible under linear pooling or in binary outcome spaces, but possible with three or more outcomes. Our framework admits recursive structure via cloning invariance, continuity, and openness, while tilt-based analysis rules out trivial duplication. Finally, we formalize an agentic alignment phenomenon in LLMs using our theory: eliciting a benevolent persona ("Luigi'") induces an antagonistic counterpart ("Waluigi"), while a manifest-then-suppress Waluigi strategy yields strictly larger first-order misalignment reduction than pure Luigi reinforcement alone. These results clarify how developing a principled mathematical framework for how subagents can coalesce into coherent higher-level entities provides novel implications for alignment in agentic AI systems.