The Long Delay to Arithmetic Generalization: When Learned Representations Outrun Behavior

Build a visualization dashboard for transformers to track the 'grokking' process by monitoring spectral entropy and model internals during algorithmic training. This helps researchers debug when a model 'knows' a task versus when it 'memorizes' it.

Suggested repo: grokwatch

"See your transformer grok the problem before it hits 100% accuracy."

Estimated effort: 20h