Grokking phase transitions in learning local rules with gradient descent
Grokking in cellular-automaton rule learning: critical exponents, grokking-probability trends, and a bimodal grokking-time distribution. Audited end to end on CPU.
5supported
0falsified
1inconclusive
0not audited
354trace events
20failures preserved
claims
C1Critical exponent in 1D exponential modelsupported
C2Critical exponent in D-dimensional uniform ball modelsupported
C3Grokking probability increases with L1 regularisationsupported
C4Grokking probability decreases with dimension Dsupported
C5Rule-30 CA learning exhibits grokkinginconclusive
C6Grokking time distribution bimodalitysupported
figure evidence
the trail
31 model calls · 199 tool runs · 39.4 min wall · append-only, failures preserved
0:00