Skip to content

Expand Solver Adapter Ecosystem #18

@fraware

Description

@fraware

Title: Implement Additional Neural Network Verification Adapters

Description:
Provability Fabric currently supports Marabou for neural network verification and DryVR for hybrid systems. The project would benefit from additional solver adapters to support a wider range of verification scenarios.

Current Adapters:
Marabou (neural networks)
DryVR (hybrid systems)

Proposed New Adapters:

  • NeuralSAT - For satisfiability-based neural network verification
  • α-β-CROWN - For complete neural network verification with improved scalability
  • Planet - For verification of neural networks with ReLU activations
  • Sherlock - For reachability analysis of neural networks

Technical Requirements:

  • Implement adapter interface following existing patterns in adapters/ directory
  • Include comprehensive test suites with benchmark comparisons
  • Provide documentation with usage examples
  • Ensure compatibility with existing specification bundle format

Expected Deliverables:

  • New adapter implementation in adapters/ directory
  • Test suite with verification benchmarks
  • Documentation updates
  • Integration tests

Skills Needed:

  • Neural network verification expertise
  • Python/Rust development
  • Understanding of formal verification concepts

Difficulty: Intermediate to Advanced

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions