Selectivity

Adapted control task metric measuring difference between odds-ratio on original task and arbitrary-label control task

Neighborhood — ranked by edge-count

framework

CausalGym
uses
Multi-task benchmark of linguistic behaviours for measuring causal efficacy of interpretability methods, adapted from SyntaxGym

method

Control task for causal evaluation
uses
Adaptation of Hewitt and Liang control tasks to CausalGym: next-token labels replaced with arbitrary tokens to measure method expressivity

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.

model selectionconcept0.787
Comparing models using log-evidence approximated by free energy.
Practicalityconcept0.783
The directness of motivation by practical concerns, characteristic of living processes in the examples.
Action Selectionconcept0.783
Choice of policies minimizing expected free energy to realize preferred future states.
Adaptabilityconcept0.782
The capacity of materials and techniques to allow fine-tuning of dimensions and shape to each unique building condition; identified as the biggest issue in achieving living architecture.
Responsivenessconcept0.773
Requirement that answers to questions be responsive as well as truthful; requires knowing that questioner will know the answer after receiving it.
Individualityconcept0.770
Formal notion of what constitutes an individual agent; bridges Buddhist and information-theoretic perspectives.
Selective attention mechanismconcept0.768
Mechanism that selects information from modules for representation in the global workspace.
Selection (Darwinian)concept0.766
The stage in the genetic algorithm where top 10% embryos by phenotypic fitness are selected.