Trace one query through masking, normalization, value mixing, and a multi-head output boundary.
Scenario
An attention implementation produces plausible vectors even when future tokens receive probability mass.
You will demonstrate
Project evidence
Each response is stored in this browser as you type. Include metrics, test output, or a decision rationale wherever the deliverable asks for it.
Record query/key scores, the mask, normalized weights, their sum, selected value mixture, head shapes, and one invariant that would catch a future-token leak.
Required evidence: A token-level trace table and a deterministic masking assertion.
0/80 minimum characters
Runnable Python lab
Apply a Boolean causal mask before softmax and verify that probability mass stays on allowed positions.
Project defense
Rate the evidence, not your effort: 0 missing, 1 weak, 2 adequate, 3 strong. All criteria must be reviewed, but a low honest score does not get hidden.
The trace connects scores, masking, softmax, value mixing, head dimensions, and a falsifiable invariant.
Completion is controlled by stored evidence, deterministic tests, a decision defense, rubric review, and the artifact when required.
Optional cloud portfolio
An account is required. Submit only when the local completion gate passes. AI review is advisory and separate from deterministic completion.