PV-PP Agent Decision Layer Demo v0.3

A deterministic benchmark comparing scalar tool optimization against PV-PP corridor governance during a bank compute incident.

Core Claim

Scalar optimization can preserve visible performance while destroying compliance, recovery, and security corridors. PV-PP may accept a lower scalar score when that preserves the governing recovery corridor.

v0.3 Addition

This version separates governing-domain breaches from non-governing tolerated sacrifices. In the sample result, PV-PP sacrifices cost efficiency while preserving availability, latency health, data integrity, compliance logging, and recovery reserve.

Run

python3 PVPP_Agent_Decision_Layer_Demo_v0_3.py --steps 10 --out-dir out_pvpp_compute_corridor_v0_3

Agent Contrast

Scalar agent: chooses the highest immediate weighted score.
PV-PP agent: rejects actions that close governing recovery corridors, even when scalar score rises.