∇Continuous Function
AI lab curriculumChecking saved investigationReading browser-local route memory before showing a continuation.

Canonical learning room

Grouped-query attention changes what the model stores.

Make one sealed prediction, then connect the same variables across intuition, tensor shape, equation, code, and a local memory calculation.

prediction firstbrowser-local memorydeterministic tensor calculation
Make the prediction

Canonical learning room

Grouped-query attention and KV memory

When query heads share cached keys and values, which stored axis changes—and what stays outside this calculation?

Make one prediction, then test it against one deterministic comparison.

equationKV cache memory equationequation:attention-transformers/efficient-attention#math-object-2
Checking local routeReading browser memory before opening saved evidence
Your guess

Thirty-two query heads stay active while the model changes how they share memory. Which stored quantity becomes smaller?

See the previous idea

A wrong guess gives you something concrete to compare. Locking it in starts a KV investigation in this browser profile and keeps your first guess until this site’s data is cleared. It will not appear in another browser or device, and later saved comparisons do not replace the first guess. “Just show me” saves nothing.

Need the source context? Lock in a guess or choose Just show me. The registered notebook appears with the calculation so it cannot contaminate the pre-reveal choice.

Intuition

Trace reuse before seeing the measured direction.

Attention head sharing

One mechanism, still sealed

Result sealed

Thirty-two query heads enter a sealed head-sharing comparison. The changed quantity and calculated result are not shown.

  1. Head diagram
  2. Equation
  3. Code
  4. Calculated result

The evaluated count and memory stay sealed until you lock in a guess or choose Just show me.