Skip to content

[Numerical] Add Cosine Similarity to Parity Reports #53

@DsThakurRawat

Description

@DsThakurRawat

Description

The numerical validation logic in src/reflex/validate.py only tracks absolute and relative differences.

Rationale

For high-dimensional action spaces (e.g., 64-dim for GR00T), directional parity is often more important than absolute distance. A model might have low L2 error but still 'flip' the direction of a joint velocity due to quantization drift. Adding Cosine Similarity provides a better metric for 'intent' parity.

Tasks

  • Implement cosine_similarity in validate_outputs.
  • Include the metric in the ValidationResult dict and CLI tables.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    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