OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Evaluation & Observability

Leanverifier

Leanverifier is a research technical preview that combines a Lean 4 library with a Python package and CLI for exact formal checks of affine binary classifiers over rational numbers.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Leanverifier — OpenViST project profile in AI Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a project involves affine binary classifiers over exact rationals and a Lean-based, command-line verification workflow. Consult its status and claims documents before relying on results because assurance is incomplete.

Who it is for

  • Researchers working with affine binary machine-learning classifiers and Lean 4.
  • Developers who need a CLI workflow to validate classifier inputs, request formal checks and inspect generated evidence.

Documented uses

  • Validate an affine binary classifier model against the CLI’s input format.
  • Request a formal check for a model and specified input, then inspect the resulting evidence.
  • Build the Lean library and run its documented axiom-audit script.
Before you choose

Requirements & limits

The repository is public and MIT-licensed. It is identified as a research technical preview; its v0.1.0 target is not authorized for release.

  • The supported model class is affine binary classifiers; neural networks, transformers and vision models are out of scope.
  • The documented robustness claim is under validation and explicitly not supported, despite implementation existing.
  • The Python CLI requires Python 3.11 or later; the Lean library requires Lean 4 matching the repository toolchain via elan.
  • Docker is optional for a Lean-only container build.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run a classifier verification request and inspect its evidence

Use the documented Python CLI flow to validate an example model, submit a verification request, and inspect its output.

Before you start

  • Python 3.11 or later.
  • A checkout of the Leanverifier repository with its example affine-binary model and request files.
  • The repository’s frozen Python dependencies, installed with uv.

Steps

  1. Install the frozen dependencies and extras with the command below.

    Documented command
    uv sync --frozen --all-extras

  2. Validate the example model with the command below.

    Documented command
    uv run leanverifier validate-model examples/affine_binary/model.json

  3. Run the command below.

    Documented command
    uv run leanverifier verify --model examples/affine_binary/model.json --request examples/affine_binary/request.json --output /tmp/lv-evidence

  4. Inspect the resulting evidence with the command below.

    Documented command
    uv run leanverifier inspect /tmp/lv-evidence

Expected outcome

The CLI produces an outcome and evidence directory that can be inspected; its exit codes distinguish verified, not-certified and several error conditions.

Guide limitations

  • The project is a research technical preview, and assurance is incomplete.
  • The documented robustness claim is under validation, not supported.
  • The supported scope is affine binary classifiers over exact rationals, not general neural networks, transformers or vision models.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
fraware/leanverifier · Official homepage title · retained in its source language
How we review sourcesSuggest a correction