← Back
Structural Motif

Phase Transition / Criticality

Look for threshold behavior where small local changes trigger regime shifts, then ask whether the AI system should detect or exploit that boundary.

Editorial Disclosure

This motif is an ISOM editorial organizing layer. It groups papers by reusable mathematical structure so transfer claims can be tested rather than presented as loose analogies.

Definition

This motif is about discontinuous or sharply nonlinear changes in macroscopic behavior as a control variable crosses a threshold.

This motif asks whether a system has a real regime boundary. ISOM requires an order parameter, a control variable, and an intervention plan before treating threshold behavior as transferable.

Mathematical Structure

The structure usually involves an order parameter, a control variable, and a connectivity or correlation story that changes the dominant regime.

Physics Side

Phase transitions and criticality explain why local couplings can suddenly become global phenomena, from percolation to collective ordering.

AI Side

In AI, analogous thresholds appear in curriculum schedules, routing sparsity, memory usage, and exploration-exploitation balance. A criticality lens can clarify where a system should stay away from instability and where it should operate near a rich transition zone.

AI candidates include curriculum switching, sparse routing, memory activation, scaling transitions, and exploration policies. The practical test is whether monitoring the boundary improves decisions compared with smooth heuristics.

Failure Modes

It is easy to call any nonlinear curve a phase transition. Without a genuine regime boundary, the metaphor only obscures a smoother control problem.

The most common failure is over-naming: many curves are nonlinear without being critical. ISOM therefore looks for discontinuity, sharp scaling, hysteresis, or a measurable change in the dominant mechanism.

Open Questions

How can we estimate usable order parameters for learning systems, and how can we intervene on them without collapsing the training dynamics?

Related Transfer Briefs

Transfer Brief

课程调度中的相变阈值

使用渗流风格的阈值估计来决定课程何时应切换模式,而不是依赖固定的时期截止。

Related Paper Analyses

Physical Review Research

Dynamical arrest in active nematic turbulence

ISOM keeps this Physical Review Research paper in the public review set because it gives readers a concrete case around Dynamical arrest in active nematic turbulence through its mechanism, assumptions, and evidence...