Passing target: 24/34 (~69%). Every code-output answer below was verified by executing the snippet on Qiskit 2.5.1 / qiskit-ibm-runtime 0.48 / qiskit-aer.

Answer Table

Q	Ans	Q	Ans	Q	Ans	Q	Ans
1	A	10	D	19	B	28	B
2	B	11	B	20	D	29	C
3	C	12	B	21	D	30	C
4	A	13	C	22	A	31	A
5	D	14	A	23	D	32	B
6	B	15	C	24	C	33	C
7	C	16	D	25	A	34	A
8	A	17	B	26	D
9	B	18	C	27	C

Explanations

Q1 — A (3). depth() ignores barriers (they are directives, not operations). The layers are: h(0) → cx(0,1) → {cx(1,2), x(0)}; the final cx and x act on disjoint qubits and share layer 3. Verified: prints 3.

Q2 — B (4). measure_all() always adds a new 2-bit register named meas (plus a barrier) — it does not reuse the existing c register unless you pass add_bits=False. So num_clbits = 2 + 2 = 4. Verified.

Q3 — C. GHZ needs one Hadamard to create superposition and a CX chain that copies the (computational-basis) value outward: h(0); cx(0,1); cx(1,2). Option B has the controls on the still-|0⟩ qubits, so the CXs do nothing; A gives an equal superposition of all 8 states; D gives |000⟩→|111⟩ deterministically (X, not H — no superposition).

Q4 — A (0). List binding follows qc.parameters, which is sorted alphabetically by name — [a, b] — not insertion order. So a=π, b=0, and rx(b) carries angle 0. Verified: the bound rx has params [0.0].

Q5 — D ({'11': 1000}). x(0) makes the first measurement deterministically 1, the if_test fires every shot and flips qubit 1, so both clbits read 1. Little-endian: clbit 0 is the rightmost character, so the key is '11' (symmetric here anyway). Verified on AerSimulator.

Q6 — B. In the {rz,sx,x} basis, H ≅ rz(π/2)·sx·rz(π/2) up to global phase. Verified: transpile produces exactly rz(π/2), sx, rz(π/2), and Operator.equiv confirms equivalence to H.

Q7 — C (1.0). Two little-endian reads at once: the label '01' puts qubit 1 in |0⟩ and qubit 0 in |1⟩; the Pauli 'ZI' puts Z on qubit 1 and I on qubit 0. Z measured on |0⟩ gives +1. Verified: prints 1.0.

Q8 — A. simplify() combines duplicate Pauli terms (XI + (−1)·XI = 0) and drops zero-coefficient terms, leaving SparsePauliOp(['IX'], coeffs=[1.+0.j]). Verified.

Q9 — B. HZH = X, so the circuit is an X gate in disguise: |0⟩ → |1⟩, giving {'1': 1.0}. If you thought the final state stayed in superposition, you forgot the second H closes the sandwich.

Q10 — D (ZIZ). from_sparse_list places Z on qubits 0 and 2 of a 3-qubit operator. In the printed string, qubit 0 is rightmost, qubit 2 leftmost → ZIZ. Verified.

Q11 — B. HXH = Z (and HZH = X). H swaps the X and Z axes of the Bloch sphere.

Q12 — B. Bell (|00⟩+|11⟩)/√2, then X on qubit 0 flips the rightmost bit of each ket: |00⟩→|01⟩, |11⟩→|10⟩ → {'01': 0.5, '10': 0.5}. Verified.

Q13 — C. V2 primitives on real hardware only accept ISA circuits (basis gates + connectivity of the target). Untranspiled abstract circuits are rejected with an error — there is no server-side auto-transpilation.

Q14 — A. In v2.x you attach a primitive to a session via mode=: SamplerV2(mode=session) inside a Session context. There is no session= kwarg, no session.run(SamplerV2), and no options.session.

Q15 — C. Iterative, feedback-dependent workloads (variational loops) are the textbook use case for Session mode — a dedicated window where successive jobs skip the public queue. Batch is for independent jobs.

Q16 — D. service.least_busy(operational=True, simulator=False) is the canonical call — it lives on the service, not the backend, and least_busy is not a kwarg of backends().

Q17 — B (4096). Shot precedence, most specific wins: PUB-level shots > run(..., shots=) > options.default_shots. The second PUB explicitly requests 4096.

Q18 — C. QuantumCircuit(2, 2) creates the implicit classical register named c, so counts live at result[0].data.c.get_counts(). .data.meas only exists when measure_all() created the meas register. Verified.

Q19 — B ('001'). Qubit 0 is 1; in the little-endian bitstring it is the rightmost character → '001'. Verified: {'001': 100}.

Q20 — D. resilience_level is an Estimator-only option (error mitigation is defined for expectation values). Sampler supports shots, dynamical decoupling, and twirling — suppression techniques.

Q21 — D ({'01': 100}). join_data() concatenates the registers' BitArrays in circuit order — alpha (holding qubit 0 = 1) stays the least-significant/rightmost bit → '01'. Verified.

Q22 — A. After transpilation the virtual qubits sit on physical qubits chosen by layout; the observable must be remapped with obs.apply_layout(isa_qc.layout) and paired with the ISA circuit. Estimator never remaps automatically.

Q23 — D ([0. 1.]). State: qubit 0 in |+⟩, qubit 1 in |0⟩. 'IZ' = Z on qubit 0 (rightmost letter) → ⟨Z⟩ on |+⟩ = 0. 'ZI' = Z on qubit 1 → ⟨Z⟩ on |0⟩ = 1. Verified: [0. 1.].

Q24 — C. Level 0 = nothing, level 1 (default) = measurement/readout mitigation (TREX), level 2 = level 1 plus ZNE. So 2 is the lowest level enabling ZNE.

Q25 — A (−1.0). ry(π)|0⟩ = |1⟩, and ⟨1|Z|1⟩ = −1. D is the reverse trap: Estimator circuits must not contain measurements, so the measurement-free circuit is correct usage. Verified.

Q26 — D. In the drawing, H is on wire q_1 and the CX control (■) is on q_1 with the target (X box) on q_0 → h(1); cx(1, 0). Top wire is always q0; don't read the diagram upside-down. Verified against draw('text').

Q27 — C. Individual qubits of a maximally entangled pair are maximally mixed; their reduced Bloch vectors have zero length, drawn at the sphere's center. The plot does not error on entangled states.

Q28 — B (0.488). Qubit 1 is the second character from the right, i.e. the left character of a 2-bit key. It is 1 only in '10' → 488/1000.

Q29 — C. The qsphere maps amplitude magnitude to blob size and relative phase to color. Bloch multivector shows per-qubit vectors (no global phase info); histograms show probabilities only.

Q30 — C. Runtime jobs are retrieved from the service: QiskitRuntimeService().job(job_id).result(). retrieve_job on a primitive class or backend is not the v2 API.

Q31 — A (−0.4). ⟨Z⟩ = (n₀ − n₁)/shots = (300 − 700)/1000 = −0.4. Sign trap: more 1s than 0s ⇒ negative.

Q32 — B. Standard error scales as 1/√shots. To cut it in half you need 4× the shots.

Q33 — C ({'11': 1000}). x q[0] sets qubit 0, cx q[0], q[1] copies it to qubit 1 — deterministic |11⟩. There is no H, so no superposition. Verified via qasm3.loads + Aer.

Q34 — A. qiskit.qasm3.dumps(qc) returns a string; dump writes to a file object, loads parses a string into a circuit, and qc.qasm() was the removed QASM-2 method.

Distractor Analysis — the 10 Trickiest

Q2 (B). Distractor A (2) is the reflex answer for people who assume measure_all() reuses the existing register; D preys on the false belief that adding a second creg is illegal. The real behavior — silently adding meas — doubles the clbits and later changes your data.<name> access path.
Q4 (A). Distractor B (π) is chosen by nearly everyone who binds by insertion order (b first). qc.parameters sorts alphabetically, so the list [π, 0] maps to (a, b). D tempts those who half-remember that dict binding is "safer" — list binding is legal.
Q5 (D). Distractor B ('10') catches readers who write the bits left-to-right as (clbit0, clbit1); C catches those who think the mid-circuit measurement outcome is random, forgetting x(0) makes it deterministic.
Q7 (C). The double little-endian trap. Distractor A (−1.0) is picked by readers who apply Z to the qubit that is in |1⟩ — i.e. who read either the state label or the Pauli string big-endian. The two rightmost characters (state 1, Pauli I) belong to qubit 0.
Q17 (B). Distractor A encodes the plausible-but-wrong mental model that options are authoritative. The precedence ladder (PUB > run call > default option) is exactly the kind of detail IBM tests.
Q18 (C). Distractor A (.data.meas) is the most common wrong answer in real life: meas exists only after measure_all(). Here the register was created by the QuantumCircuit(2, 2) shorthand, so its name is c. B is the retired V1 (quasi_dists-era) access style.
Q21 (D). Distractors B ('0 1') mimics the space-separated multi-register formatting of the old get_counts() on V1 results; C tempts those who think join_data errors on multiple registers when joining them is precisely its purpose. Register order keeps alpha on the right.
Q23 (D). Distractor A ([1. 0.]) is the big-endian reading of IZ/ZI. If you can articulate why 'IZ' touches qubit 0, you've internalized the single most-tested convention on this exam.
Q24 (C). Distractor B (1) catches candidates who remember "level 1 is the default and does mitigation" and stop reading — level 1 is TREX only; ZNE arrives at level 2.
Q26 (D). Distractors A and B both come from misreading which wire is q0 (top) or which end of the CX is the control. The box (X) is the target on q_0; the filled dot on q_1 is the control — so the control qubit index comes first in cx(1, 0).