method
active
method:intervention-validationIntervention validation
Perturbing source features along decoder direction and measuring target change.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Attribution graphs with nodes per turn per layer.
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.
- CIMC's methodology for evaluating whether a built system is conscious: combining multiple forms of evidence including predicted functional organization and developmental trajectories
- Number of latent variables assigned per algorithm node in distributed abstraction; affects IIA
- Property that additive modifications to activations affect all downstream computations, enabling tractable behavioral control
- Scalar parameter modulating how strongly a steering vector shifts model activations; set to 15 for Exp1 and ±16 for Exp2
- Fundamental operation for causal abstraction analysis; forces neurons to take values from source inputs to create counterfactuals.
- Intervention targeting specific dimensional subsets of activation vectors rather than full representations
- The use of interventions (rather than correlations) to establish a causal link between representation geometry and behavioral geometry.
- Proportion of aligned interchange interventions with equivalent high-level and low-level effects; graded measure of causal abstraction.