Input-Injectivity

Assumption that DNN layers preserve input information by being injective; key condition for Theorem 1

Neighborhood — ranked by edge-count

paper

finding

concept

Neural Collapse
contradicts
Terminal phase phenomenon in deep learning training relevant to convergence of representations

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.

Input-truthconcept0.789
Correctness of input statements to an LLM, as opposed to output-truth (correctness of model-generated outputs).
sensory inputconcept0.781
Input from environment that the agent models and predicts.
Distinguishing Injected Concepts from Text Inputsfinding0.778
Models maintain ability to accurately transcribe input text while simultaneously reporting on injected thoughts, all models perform above chance, Opus 4/4.1 best.
Input-Restricted Interventionconcept0.763
Practical restriction of interventions to those producible by actual inputs; standard in DAS practice
Inventiveness and Creativityconcept0.758
The cognitive outcomes that are directly proportional to the presence of loose parts, according to Nicholson.
Input-Output Relations as Ordered Conceptsfinding0.748
Diagrammatic encoding of program behavior via concept lattices reveals reachability structure and non-determinism without fixed calculational rules.
Input-Output Specificationconcept0.742
Specification relating a program's inputs and outputs, analogous to illocutionary correctness.
Interactionconcept0.742