Asynchronous Update Training

Training regime where random subsets of cells update per step, improving robustness of learned circuits

Neighborhood — ranked by edge-count

framework

Differentiable Logic Cellular Automata (DiffLogic CA)
uses
The novel framework introduced in this paper, combining DLGN and NCA for fully differentiable discrete CA learning

concept

Asynchronous Updates
implements
Update strategy where random subsets of cells are updated per step, simulating independent cell clocks

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.

Models trained directly with asynchronous updates would exhibit even greater robustness than synchronously trained modelshypothesis0.788
Hypothesis that motivated the asynchronous robustness comparison experiment
asynchronous communicationconcept0.775
Communication where sender does not block; dominant in Linda.
Synchronous Updatesconcept0.770
Default update strategy where all cells update simultaneously; contrasted with asynchronous updates
Asynchronous training improves robustness of DiffLogic CA circuits to perturbations compared to synchronous trainingclaim0.749
Interpretive claim supported by damage resilience experiment results
Character Trainingconcept0.726
The post-training approach used by frontier AI labs to shape the assistant persona, introduced as open-source in this paper
Harness-Updating Capabilityconcept0.721
The capability of an evolver model to produce useful persistent harness updates from execution evidence
Pretrainingconcept0.716
Initial large-scale training phase whose early stages are shown to form persona representations
Associative Learningconcept0.711
Sentience criterion; capacity occurs even in gene regulatory networks and non-neural morphogenetic agents.