IBM C1000-179 — Fundamentals of Quantum Computing Using Qiskit v2.X Developer Passing target: 24/34. All code outputs were verified by execution with qiskit 2.5.1 / qiskit-ibm-runtime 0.48.0 / qiskit-aer.

Answer Key

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

Explanations

Q1 — B (['cx', 'rz', 'sx']). Translation to the basis set rewrites h as rz–sx–rz, while cx is already a basis gate and survives. Verified by execution. Distractor analysis (tricky #1): A assumes h is left alone because no backend was given — but basis_gates alone triggers translation; C forgets the CX is kept; D is the old U-gate basis from Qiskit 0.x.

Q2 — B (4). Layer 0: the three H gates. Layer 1: cx(0,1). Layer 2: cx(1,2) (waits on qubit 1). Layer 3: cz(0,2) — qubit 0 has been free since layer 1, but qubit 2 is busy until layer 2, so the CZ starts a fourth layer. Verified by execution.

Q3 — A (0). Optimization level 0 still performs layout, routing, and translation — it only skips the optimization stage. Every h is translated into rz/sx, so count_ops().get('h', 0) is 0 (the actual ops are {'rz': 10, 'sx': 5, 'cx': 5}). Verified by execution. Distractor analysis (tricky #2): B/C assume level 0 leaves gates untranslated — the classic "level 0 = do nothing" myth; D is wrong because level 0 does route (it must, to satisfy the coupling map).

Q4 — B. The preset staged pass manager runs init → layout → routing → translation → optimization → scheduling. Remember: you must pick where qubits go (layout) and insert SWAPs (routing) before rewriting to basis gates (translation) makes final optimization meaningful.

Q5 — A (4 4). compose simply concatenates the instructions of b = a.inverse() after a; nothing is cancelled automatically, so size = 2 + 2 = 4 and the gates stack into 4 layers. Verified by execution. Distractor analysis (tricky #3): B is the trap — a circuit composed with its inverse is the identity operator, but Qiskit does not simplify unless you transpile with optimization; C miscounts depth.

Q6 — B. Level 3 applies the heaviest optimization passes and generally produces the shallowest ISA circuits. A is false (level 0 still does layout/routing), C is false (pm.run output is already ISA), and D is false (ISA is required at every level).

Q7 — B ({'10': 1.0}). x(1) flips qubit 1. Little-endian: qubit 1 is the second character from the right, so the state is '10'. Distractor analysis (tricky #4): A ('01') is the big-endian misreading — flipping "the second qubit" and writing it second from the left. Verified by execution.

Q8 — B (−1.0). from_label('01') reads left→right as q1, q0, so q1 = |0⟩ and q0 = |1⟩. ZZ contributes (+1)·(−1) = −1. expectation_value accepts SparsePauliOp fine (D is false). Verified by execution.

Q9 — A ({'0': 1.0}). H·X·H = Z (the conjugation identity), and Z|0⟩ = |0⟩ — the phase flip is invisible on |0⟩. Verified by execution. Distractor analysis (tricky #5): B is the trap for "there's an X in there, so it must flip"; C is for treating the H's as independent randomizers; D forgets statevector simulation is deterministic.

Q10 — A (True). Three alternating CNOTs implement SWAP exactly — a standard identity. Operator.equiv compares unitaries up to global phase, and no transpilation is needed to build an Operator (C, D false). Verified by execution.

Q11 — A. SparsePauliOp.from_list takes (pauli_string, coefficient) tuples in that order. B reverses the tuple order, C is not a valid Pauli constructor, and D mixes strings into arithmetic and does not parse.

Q13 — C (Batch mode). The circuits are independent, there is no feedback loop, and batch mode lets the scheduler pack the jobs efficiently without the exclusive reservation (and idle-time cost) of a session. Distractor analysis (tricky #6): B (Session) is the reflex answer for "many circuits", but sessions are for iterative workloads where job N+1 depends on job N; A works but forfeits batch's parallel scheduling; D cannot model hardware noise for benchmarking real devices.

Q14 — A. Open the session as a context manager and pass the session object via mode=. B passes a string, C never links the session to the primitive, and D creates the primitive without any mode, so it does not run inside the session.

Q15 — B. V2 primitives on IBM hardware accept only ISA circuits — circuits already expressed in the target's basis gates and respecting its connectivity. Non-ISA circuits are rejected with an error; the service does not transpile for you (A is the old V1-era behavior).

Q16 — A. Job mode is the simplest mode: one primitive request, one independently queued job, no dedicated window and no state carried between jobs.

Q17 — A ({'1': 100}). The default classical register is named c, so counts live at result[0].data.c. X|0⟩ = |1⟩ deterministically, giving all 100 shots on '1'. C is the V1 quasi-distribution habit; local samplers do accept shots (D false). Verified by execution.

Q18 — A ({'01': 200}). measure_all() creates a register named meas, so .data.meas is correct (C false). x(0) flips qubit 0 = rightmost bit → '01'. B is the big-endian trap. Verified by execution.

Q19 — C. The Sampler samples classical bits; with no measurements there is nothing to sample, and the run errors out. It never invents measurements for you (D) and never returns statevectors (A).

Q20 — A. Dynamical decoupling lives under options.dynamical_decoupling with enable and sequence_type ("XX", "XpXm", "XY4"). B is an Estimator-only option, C is not a real attribute, D enables twirling — a different suppression technique.

Q21 — D. resilience_level (and the ZNE/PEC machinery behind it) belongs to EstimatorV2 only. Sampler supports shots, dynamical decoupling, and twirling. Distractor analysis (tricky #7): B and C tempt because DD/twirling sound like "mitigation" that Sampler shouldn't have — but they are suppression options and are available on both primitives.

Q22 — A ([1. 1. 0.]). For the Bell state (|00⟩+|11⟩)/√2: ⟨ZZ⟩ = 1 (bits always agree), ⟨XX⟩ = +1 (it is an XX eigenstate), ⟨XZ⟩ = 0. One PUB with a list of observables broadcasts into an array of expectation values. Verified by execution. Distractor analysis (tricky #8): B flips the XX sign — that would be the (|01⟩+|10⟩)/√2 state under ZZ, a mixed-up memory of which Bell state is which.

Q23 — A. After transpilation the virtual qubits sit on physical qubits chosen by layout, so the observable must be expanded/permuted with obs.apply_layout(isa.layout). The layout lives on the transpiled circuit (B uses the untranspiled one, which has no layout), and the service does not do this for you (D).

Q24 — B. Level 1 = measurement (readout) error mitigation via TREX; level 2 = level 1 plus zero-noise extrapolation. PEC is not part of any resilience level preset in this exam's scope — it is enabled explicitly; DD is suppression, not a resilience level.

Q25 — A, C. ZNE and PEC are mitigation: classical post-processing that estimates the zero-noise answer after execution. Dynamical decoupling and Pauli twirling are suppression: pulse/gate-level interventions applied before or during execution. Distractor analysis (tricky #9): twirling (D) is the most common wrong pick — it randomizes noise to make it Pauli-like during execution, which is suppression even though it helps mitigation work better.

Q26 — B. h(1) puts H on the q_1 wire (bottom — q0 is always the top wire), and cx(1, 0) puts the control dot on q_1 and the ⊕ target on q_0. A draws h(0); cx(0,1), C draws h(1); cx(0,1), D draws h(0); cx(1,0). Diagrams generated with qiskit 2.5.1.

Q27 — C. Each qubit of a maximally entangled pair is locally the maximally mixed state, whose Bloch vector has zero length — both arrows collapse to the center. This is the visual signature of entanglement in plot_bloch_multivector.

Q28 — A. plot_histogram accepts a list of counts dictionaries plus a legend list of the same length. B tries to add dicts (TypeError), C merges them so hardware values overwrite simulator values, D passes the second dict into the wrong positional argument.

Q29 — B. On the qsphere, blob size encodes the probability/amplitude magnitude and color encodes the relative phase of each basis-state amplitude. That is what distinguishes it from a plain histogram.

Q30 — A (0.2). ⟨Z⟩ = (n₀ − n₁)/shots = (600 − 400)/1000 = 0.2. B has the sign flipped (Z assigns +1 to |0⟩, not to |1⟩); C is just p(0); D is p(0)·p(1)·... noise.

Q31 — A. service.job(job_id) retrieves a single job by ID in qiskit-ibm-runtime. service.jobs() is for filtered listings (and takes filter kwargs, not a bare ID); the other calls do not exist.

Q32 — B (4,000). Standard error scales as 1/√shots. Halving the error requires 4× the shots: 1,000 → 4,000. Distractor analysis (tricky #10): A (2,000) is the linear-scaling trap — doubling shots only reduces error by a factor of √2 ≈ 1.41.

Q33 — A (2 2). c = measure q; is valid OpenQASM 3 register-wide measurement syntax; the importer expands it into one measure instruction per qubit, so count_ops()['measure'] is 2, and the circuit has 2 qubits. Verified by execution.

Q34 — B. OpenQASM 3 adds real classical control flow (if, while, for), typed classical data, and input parameters. include, quantum registers, and cx all exist in OpenQASM 2 already.

Ten Trickiest — Quick Index

#	Q	Trap theme
1	Q1	Thinking translation needs a backend object
2	Q3	"Optimization level 0 does nothing" myth
3	Q5	Expecting compose(inverse) to auto-cancel
4	Q7	Big-endian bitstring misreading
5	Q9	Missing the HXH = Z conjugation identity
6	Q13	Session-vs-batch mode confusion
7	Q21	DD/twirling wrongly assumed Estimator-only
8	Q22	Bell-state ⟨XX⟩ sign confusion
9	Q25	Twirling misfiled as post-processing mitigation
10	Q32	Linear (not quadratic) shot-scaling error