Skip to content

Latest commit

 

History

History
48 lines (28 loc) · 3.06 KB

File metadata and controls

48 lines (28 loc) · 3.06 KB

exp-depth-params — Is Grassmannian's advantage geometric or just more parameters?

Round 6 · Confirming the mechanism


What we were asking

Grassmannian attention uses a k×k inner product matrix per attention pair, which means — naively — it has more parameters in the Q and K projections than standard attention at the same d_head. Could the depth-6 advantage simply be explained by having more parameters, rather than by the geometric properties of subspace comparison?

This is a critical sanity check. Science requires ruling out simpler explanations before claiming a novel geometric effect.

The analogy

Suppose a larger-engined car beats a smaller-engined car in a race. You can't conclude that engine design matters until you also compare two cars with the same engine size but different designs. This experiment is that comparison: same Q/K parameter count, different similarity primitive.

What was tested

At tree depth 6 (the phase transition point), we ran three variants:

  1. Standard attention — d_head=16, Q/K params = 16,384
  2. Grassmannian k=2, parameter-matched — d_head=8 for Q/K, Q/K params = 16,384 (identical to standard)
  3. Grassmannian k=2, extra params — d_head=16, Q/K params = 32,768 (2× standard)

Results

Model Q/K params Total params Final accuracy (5K steps)
Standard attention 16,384 107,872 23.9%
Grassmannian (matched) 16,384 99,680 97.9%
Grassmannian (extra) 32,768 124,256 98.5%

The parameter-matched Grassmannian — using fewer total parameters than standard attention — achieves 97.9% vs 23.9%. The advantage is geometric, not parametric.

What this means

The subspace overlap primitive provides fundamentally more geometric expressiveness than the dot-product primitive at equal parameter cost. With the same number of weights to learn, the k=2 Grassmannian similarity can solve a problem that k=1 dot-product cannot.

The intuition: with 64 leaves and 64 dimensions, each token must simultaneously "point toward" all its tree-neighbours and "point away from" all non-neighbours. A 1D projection (dot-product) can satisfy one such constraint at a time. A 2D subspace projection has the geometric degrees of freedom to satisfy more constraints simultaneously — enough to solve the packing problem at d=N.

The parameter-matched Grassmannian also has fewer total parameters (because d_head=8 rather than 16 for the value and output projections), yet outperforms standard attention dramatically. This is a clean geometric story: better primitive, same budget, much better result.

The extra-param Grassmannian is only marginally better (98.5% vs 97.9%), confirming the advantage comes from the k=2 subspace structure, not from scale.

Open questions

  • Does the parameter-matched Grassmannian also succeed at depth 7? Or does it share standard attention's failure at d/N < 1?
  • At what d/N ratio does even Grassmannian fail — what is the geometric expressiveness limit of k=2 subspace comparison?