QuantumComputingBench · graph-grounded exercise book
The same five-part curriculum, with no duplicate mathematical core counting. A source-supported NL answer is not automatically a verified formal view. Admitted views and remaining formalization work are labeled separately.
3 admitted native Lean views in this checkpoint. These check exact conclusions by library-lemma application, not deep proof synthesis or the full NL explanation. The three corrected-v3 reference warmups are separate legacy diagnostics, not additions to the native count. Trusted local checking does not establish isolated model-output evaluation.
Open the three semantic-audited API warmups
Replacing two distinct rows i and j of a square matrix by the linear mixtures (a*row_i + b*row_j) and (c*row_i + d*row_j) multiplies the determinant by (a*d - b*c). Use this to prove that a plane rotation acting on rows …
local source supported
For a finite real amplitude table f on n+1 wires, define its low-wire marginal by taking, at each prefix word b, the Euclidean length of the sibling pair (f at b-with-last-bit-0, f at b-with-last-bit-1). Prove that the t…
nl bridge from inspected source
A real tensor-train core carries orthonormal rows (its row-canonical condition). Show that embedding such a core into a complex matrix with the physical bit and new bond as output rows and the old bond as input columns y…
nl bridge from inspected source
Start with dimensions, inner products, isometries and explicit conventions. The same foundations support state preparation, block encoding and later quantum-information arguments.
Reason from a target amplitude family to normalization, structural representations and a preparation construction. Inspect sparse and zero boundaries before transferring an argument to a circuit interface.
Track the represented operator, its normalizer and the accepted subspace. Keep preparation, SELECT, unprepare, workspace cleanup and resource tiers visible when composing constructions.
Study state and channel representations, symmetry and tensor structure through primary-source-grounded problems. Planned curriculum anchors are not local Lean closure for the complete answers.
Follow the roles of input states, operator access, polynomial transformations and output interpretation. Source-only algorithm questions keep their assumptions and query/gate/readout boundaries explicit.