Interpretability and mechanistic analysis · MLD-1188
Superposition
Representation of more features than available dimensions through overlapping, nonorthogonal directions under a modeling hypothesis.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Study sparse features packed into a residual stream.
Original caveat
Not every distributed representation demonstrates superposition.
Context and aliases
Level: Advanced. Stages: Research · Evaluation.
Tags: ML · DL · SLM · LLM · Interpretability · Safety.
Aliases: None recorded.