Distinguish or Homogenize: Last-Chance Policy Identification and Risk-Budgeted Recovery under Irreversible Resource Depletion

2026-09-29

AI Summary

This study addresses the dilemma agents face under irreversible resource depletion: whether to diagnose fault models or alter system states to accommodate a unified policy. To resolve this, it proposes a “last-chance policy identification” framework that establishes a “distinguish-or-homogenize” principle, transcending traditional static mapping limitations by modeling acceptable policy mappings as action functions and defining identifiable boundaries for recovery planning under risk budgets. The approach is validated through deterministic diagnostic graph recursion, risk-budget-compatible planning, and pseudo-control experiments. Results demonstrate that the framework significantly improves the rate of risk-feasible recovery in both microservice topologies and MiniGrid environments while satisfying failure budgets, confirming that performance gains stem from policy compatibility transformations rather than additional search overhead.

Abstract

Under irreversible resource depletion, an agent can spend resources to distinguish among latent fault models, or to change the system state so that the remaining models admit a common acceptable continuation--at which point further diagnosis becomes unnecessary. This distinguish-or-homogenize principle identifies a path that existing frameworks for identification, planning, and diagnosis do not make explicit: prior formulations treat the mapping from fault models to acceptable policies as a given, whereas LCPI makes it a function of the agent's own actions. We formalize this principle through Last-Chance Policy Identification (LCPI), where correctness is evaluated at the state the agent reaches rather than at the initial state. The Last Identifiable Margin (LIM) marks the feasibility boundary between distinguishing and homogenizing. For deterministic diagnostic graphs we provide the Exact-LIM recursion; for noisy finite-horizon recovery we propose Risk-Budgeted Compatibility Planning (RBCP), which searches a compatibility-aware frontier under a hard worst-case failure constraint. Across incident recovery on abstract microservice topologies and latent-damage navigation in MiniGrid, RBCP improves risk-feasible recovery while satisfying the failure budget. A sham control--cost-matched actions that preserve model incompatibility--eliminates the gain entirely, confirming that the benefit comes from changing which policies are acceptable for which models, not from extra search or additional budget.

Problem

• irreversible resource depletion
• fault model identification
• policy homogenization
• incident recovery
• risk-budgeted planning

Innovation

• Last-Chance Policy Identification
• Risk-Budgeted Compatibility Planning
• Last Identifiable Margin
• Irreversible Resource Depletion
• Distinguish-or-Homogenize