Deriva — Derive the Algorithm

← ProjectBuild Everything / CAP2Novel Attention + Paper~2160 min

CAP2test a new hypothesis400 lines · M3 + research

Test a sparse-attention hypothesis with a fair baseline: each position averages only the most recent window tokens.

Required API

def sparse_attention(values, window):
    return outputs

Behavior

Optional design question

Think through the mechanism first if you want the extra reasoning step. It never blocks the editor.

Visible contract tests

•
window twosparse_attention([1.0, 3.0, 5.0], 2)
sparsity changes context width, not output length
•
window onesparse_attention([1.0, 3.0], 1)
the smallest sparse window is local identity

2 hidden edge tests run after the visible contract passes.

def sparse_attention(values, window):
    # Use a causal trailing window of at most `window` values.
    pass

Run Tests to see the contract verdicts here.

0 runs

Opening…