method
active
method:ravel-evaluationRAVEL evaluation
Disentanglement metric via latent interventions with Cause and Isolation metrics.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- SAEBenchusesEvaluation suite of eight metrics across four capabilities.
Concepts (1)
concept
- Independent concepts are encoded by separate latents.
Methods (2)
method
- Used to select latents in RAVEL.
- Skyline for RAVEL achieving 0.87.
Datasets (1)
dataset
- RAVEL datasetusesCities and Nobel laureates with attributes.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- A specific signal (Wood Labs) embedded in evaluation environments that the model organism uses to reliably identify testing contexts.
- Nielsen and Molich's method for finding UI flaws by applying usability heuristics.
- Using Claude Sonnet 4 as a grader to categorize model responses according to predefined criteria.
- Evaluation setting where the same task stream that drives evolution also serves as the evaluation set, with each task scored under the harness at time of attempt
- Agent that plans, implements, and revises behavioral evaluations using Claude Code.
- Mental states that guide behaviour via assessments of what is good, right, or rational.
- nostalgebraist's term for measuring performance when the model is incentivised to perform well.