https://github.com/z139-lab/uvp-v2-framework
"Universal scaling analysis for bounded criticality in multi-agent systems."
https://github.com/z139-lab/uvp-v2-framework
bounded-criticality complex-systems criticality multi-agent reproducibility
Last synced: about 2 months ago
JSON representation
"Universal scaling analysis for bounded criticality in multi-agent systems."
- Host: GitHub
- URL: https://github.com/z139-lab/uvp-v2-framework
- Owner: Z139-Lab
- License: other
- Created: 2026-04-19T15:43:40.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2026-04-21T18:42:34.000Z (2 months ago)
- Last Synced: 2026-04-21T20:42:29.328Z (2 months ago)
- Topics: bounded-criticality, complex-systems, criticality, multi-agent, reproducibility
- Language: Python
- Homepage:
- Size: 778 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Citation: CITATION.cff
Awesome Lists containing this project
README
\# UVP v2: Universality Validation Pipeline for Bounded Criticality
A reproducible framework for detecting, comparing, and analyzing bounded criticality across complex systems.
\---
\## 🔁 Reproduce Main Results (One Command)
```bash
pip install -r requirements.txt
python run\\\_uvp.py --config configs/grid\\\_default.yaml
Outputs:
Critical region detection
Collapse-style analysis
Dynamical exponent estimation (z)
Publication-ready figures (saved in results/)
Overview
UVP v2 (Universality Validation Pipeline v2) is a unified analysis framework designed to test whether different complex systems exhibit a shared form of bounded criticality.
The pipeline transforms heterogeneous simulation outputs into a standardized statistical-physics workflow, enabling:
critical-region detection
collapse-style comparison
dynamical exponent estimation
cross-system universality testing
UVP v2 is designed for use across:
power-grid cascading-failure systems
LLM / AI multi-agent cascade systems
Core Idea
Many complex systems do not exhibit purely divergent criticality.
Instead, they display a bounded critical regime:
a transition region with identifiable critical structure, but without unconstrained divergence.
UVP v2 provides a systematic framework to detect and analyze this regime.
Pipeline
raw data
→ normalization
→ stress scan
→ critical detection
→ collapse analysis
→ z estimation
→ publication-ready outputs
Key Functions
Unified data adapters for heterogeneous simulation outputs
Critical region detection using configurable heuristics
Collapse-style analysis across system sizes or variants
Dynamical exponent estimation (proxy-based and tau-based)
End-to-end reproducibility from raw data to figures
Cross-system comparison (infrastructure systems vs AI agent systems)
Input Schema (Canonical Representation)
All datasets are mapped to a unified schema:
stress
L (system size)
phi (observable)
collapse\\\_prob (optional)
tau (optional)
n\\\_seeds (optional)
Additional metadata (e.g., chi, cv, variant, topology) are preserved.
Expected Outputs
Each run produces:
processed\\\_input.csv
sigma\\\_scan/scan.csv
critical/sigma\\\_c.json
collapse/collapse\\\_data.csv
z\\\_fit/z\\\_fit.json
figures/scan.png
figures/collapse.png
figures/z\\\_scan.png
Scientific Scope
UVP v2 is designed to investigate:
finite-width critical bands vs classical phase transitions
robustness under micro-rule and topology perturbations
cross-system universality of bounded criticality
scaling behavior across heterogeneous systems
Current Status
This repository provides a framework layer, not a finalized claim engine.
Limitations:
critical detection is heuristic-based
z estimation depends on data richness
collapse quality is not yet fully optimized
Typical Workflow
raw simulation data
→ standardized input format
→ critical region detection
→ collapse analysis
→ exponent estimation
→ publication-ready summaries
Relation to Other Repositories
Power-grid application:
https://github.com/Z139-Lab/bounded-criticality-power-grids
Research portal:
https://github.com/Z139-Lab/bounded-criticality-portal
IEEE-2383 case study:
https://github.com/Z139-Lab/ieee-2383-critical-edge
Recommended Use
Use UVP v2 when you want to:
compare bounded-critical behavior across systems
distinguish sharp vs finite-width transitions
estimate dynamical scaling behavior
build reproducible, publication-grade analysis pipelines
Reproducibility
All results are reproducible via:
python run\\\_uvp.py --config configs/grid\\\_default.yaml
The pipeline produces deterministic outputs given fixed seeds and configuration.
Citation
Please use the included CITATION.cff file for formal citation.
Author
Juan Adam
Independent Research