method
active
method:concreteness-judge

Concreteness Judge

LLM-based judge rating SAE latent labels 0-100 for concreteness to filter steering candidates

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

  • Validation of judge model robustness by regrading 1000 responses with 4 additional judge models
  • Pre-filtering step excluding abstract latents where off-topic detection is harder
  • Modular reasoning enabled by monoid and other algebraic laws; central to maintainability and correctness of functional systems.
  • Using Claude Sonnet 4 as a grader to categorize model responses according to predefined criteria.
  • Contrastconcept0.700
    The property that living structures contain intense contrast—far more than one imagines helpful; true opposites which annihilate each other when superimposed, creating differentiation that gives birth to something; contrast unifies rather than separates when used correctly
  • Competencyconcept0.699
    Ability of cells or tissues to actively restore correct shapes despite perturbations and barriers.
  • Claude 4.5 Haiku used to segment responses into attempts and score each attempt 0-100 for relevance
  • An LLM-based classifier that returns 1 if response contains a clear subjective experience report and 0 otherwise