mrv-lib: Model Risk Validator

Getting started

• Installation
  • Requirements
  • Standard install
  • Development install
  • Building documentation
• Quick start
  • Labels-first API
  • Validation report
  • Next steps

Tutorials

• Tutorial: Paper 1 – Representation Invariance
  • Theory background
  • Step 1: prepare labels
  • Step 2: run the validator
  • The 1/K null
  • Ordering consistency
  • Interpreting results
  • See also
• Tutorial: Paper 2 – Resolution Invariance
  • Theory background
  • Step 1: build per-frequency label dicts
  • Step 2: run the validator
  • Within-intraday excess
  • Permutation p-values
  • Using PAPER2_FREQS
  • Interpreting results
  • See also

API reference

• mrv (top-level)
  • Subpackages
• mrv.invariance – High-level invariance API
  • Sub-modules
• mrv.validator – Regime diagnostics
  • Validators
  • BaseValidator
  • RepValidator
  • ResValidator
  • generate_report()
  • Sub-modules
• mrv.pipeline – Labels-first validation pipeline
  • validate_rep()
  • validate_res()
  • report()
  • download()
  • run()
  • validate()
• mrv.models – Regime model fitting
  • fit()
  • register_model()
  • Sub-modules
• mrv.utils – Configuration and logging
  • load()
  • get_data_dir()
  • get_assets()
  • setup_logging()
  • Sub-modules

Project

• Changelog

mrv-lib is a pure Model Risk Validator library: you supply labels from your own models and mrv measures how stable they are across specification choices.

• Representation invariance (Paper 1): do regime labels change when you switch feature representations?

• Resolution invariance (Paper 2): do labels agree across 5m/15m/1h/1d frequencies?

The validation output supports OCC Bulletin 2026-13 / SR 26-2 (supersedes SR 11-7 from 2026-04-17) model risk management workflows.

Indices

• Index

• Module Index

• Search Page