Sampling-Based Quantum Diagonalization | Anupama Ray, IBM Research

Value of the Information & Strength of the Argument

The value of the information is substantial for researchers working on quantum chemistry algorithms. It offers a clear mapping of the algorithmic landscape (SQD, SKQD, SQDrift) and positions DF‑SQD as a practical compromise, achieving shallow circuits without losing the sampling‑based advantage. The argumentation is solid: each design choice in DF‑SQD is justified—e.g., selecting double‑factorized leaves for their configuration‑generating power rather than mere reconstruction accuracy, and seeding with pair amplitudes to avoid blind sampling. The talk gives explicit reasons for convergence behavior (finite support recovery, nested space monotonicity) and backs claims with numerical results. The speaker also honestly discusses limitations, such as the absence of formal convergence proofs and the continued dependency on an appropriate ansatz, albeit a more generic one. Overall, the reasoning is coherent and clear, though some parts (like the detailed derivation of the gate blocks) are sketched briefly to stay within the seminar format.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high. The work is drawn from a recent arXiv preprint (2609.01264) and reported by the lead researcher herself, providing first‑hand access to the methodology. The talk references established concepts (configuration interaction, Krylov space, QDrift) and the comparisons with SKQD and SQDrift are well‑grounded. The only formal source explicitly cited is the arXiv paper, which is sufficient given the format. The title and content are well aligned: the presentation covers the entire SQD family and introduces a new variant, so it is not overstated. The use of specific examples (iron‑sulfur cluster, methane dimer, 12,000‑atom protein) adds credibility, but these are anecdotal and not explored in depth. No external validation or independent sources are provided, which is expected for a seminar talk, but it limits the ability to independently verify the claims without further reading.

The title accurately reflects the content: the talk details the sampling-based quantum diagonalization family and introduces the DF-SQD variant. The presentation covers the core methodology, motivations, results, and comparisons, fully matching the title.

8/10

Presentation by a senior IBM research scientist of her own recent peer-reviewed work (arXiv:2609.01264). Technical depth is high, results from both simulators and quantum hardware are shown, and the algorithm is positioned against existing methods. Some claims (e.g., scalability) are illustrated but not fully proven in the talk, which prevents a perfect score.

• 00:00Welcome and speaker introduction
• 03:38Talk overview and motivation
• 06:10Why sampling-based quantum diagonalization?
• 11:35How the SQD workflow works
• 16:51SQD and Krylov variants: Trade-offs
• 23:16Introducing deterministic fields for SQD
• 24:50DF-SQD algorithm step by step
• 34:24Simulator and hardware results
• 45:35SQD Agent: Guided quantum chemistry workflows
• 54:04Questions and discussion

• Deterministic Fields for Sampling‑Based Quantum Diagonalization (arXiv) — The research paper presenting the DF‑SQD algorithm, introduced and discussed by the speaker.
• QuantumGrad Qiskit Fall Fest 2026 — The event page for the seminar series, providing context and registration details.
• QuantumGrad Qiskit Fall Fest Hackathon — Information about the hackathon, including IQM hardware credits, mentioned in the description.
• QuantumGrad Seminar Series Playlist — Playlist of the seminar series where this talk is featured.

• Deterministic Fields for Sampling‑Based Quantum Diagonalization (arXiv) — The paper itself, which the talk is based on, serving as the primary source of the algorithm and results.