question
active
question:what-is-the-correct-formal-definition-of-a-copying-matrix-that-captures-all-and-only-the-cases-we-care-aboutWhat is the correct formal definition of a 'copying matrix' that captures all and only the cases we care about?
Open methodological question about summarizing OV matrix behavior; eigenvalues are used as a working but imperfect proxy
Neighborhood — ranked by edge-count
Papers (1)
paper
- A Mathematical Framework for Transformer Circuitsassociated_with
Claims (1)
claim
- Empirical observation from examining expanded OV/QK matrices; approximately 10 out of 12 heads show significant copying
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.
- An OV circuit matrix that maps tokens to increasing the logit of those same tokens; detectable via positive eigenvalues
- A warning against oversimplifying the nature of life-quality.
- Ganter's assertion that concept lattice transformation does not lose information from the original formal context.
- Freud's insight on dynamic memory rearrangement.
- What is an appropriate definition of machine, and does it apply to all, some, or no living forms?question0.726Central interrogative driving the paper's entire argumentative structure.
- Mere copying of tokens between paired positions suffices to simulate all partial recursive functions and model higher-order logics.
- Critique of using formal specifications alone for concept definition.
- The Geometry of Interaction model shows that simple copying of information between locations suffices for all computation, establishing emergent logic.