{"id":"ea181089-6fc0-49b8-9156-da193835a00a","arxiv_id":"2504.16579","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":2,"one_line_summary":"A framework that replaces statically known mid-circuit measurements and resets with probabilistic or static subcircuits, controlled by a tunable cost knob n_pcm, whose reset extension is invalid for entangled states.","lead":"This paper extends a prior probabilistic circuit model so quantum compilers can replace known mid-circuit measurements and resets with cheaper fixed subcircuits. It reports large reductions on random dynamic circuits and in a qubit-reuse example, but the reset result rests on a wrong model of what resetting an entangled qubit does.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2's reset replacement models reset as postselection, not as the CPTP reset used on hardware; for entangled inputs the replacement is not state-equivalent and the reset-removal results are unsupported.","rationale":"The reader's weakest assumption is exactly the load-bearing flaw. Section 3.3's abstract reset rule and Theorem 2 both identify reset with normalized projection. For any entangled input, the concrete reset is a CPTP map producing a mixed state on the remaining qubits, and no unitary replacement can be equivalent. This makes the reset-removal contribution unsound as stated, so the REJECT verdict is justified. I am not raising a separate objection: the measurement-reduction and QCP-separation material are useful, and the paper has a public implementation, but those do not rescue the central reset claim. The proposed Bell-state check is minimal and would settle the issue by direct comparison; if the optimized circuit happens to match the mixed-state statistics, that would indicate the implementation does not actually follow Theorem 2, so the theorem still needs revision. Verdict unchanged.","tokens_in":11600,"tokens_out":6309,"duration_ms":67601,"concrete_test":"Run the public repository's optimizer on the 2-qubit circuit: H on q0, CNOT q0->q1 (preparing (|00⟩+|11⟩)/√2), then a reset of q0, then measure q1. Compare (a) the simulator's output distribution under a standard reset, and (b) the distribution produced by the Algorithm 1/Theorem 2-optimized circuit. A correct reset gives Pr(q1=0)=Pr(q1=1)=1/2 in the Z basis, and likewise in X, because the reduced state is I/2. The paper's unitary replacement gives a pure state on q1, so the distribution differs. If the code instead reproduces the mixed statistics, inspect the generated circuit: it must have inserted a measurement or ancilla, contradicting Theorem 2's static-unitary claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reset branch of Algorithm 1 rests on Theorem 2, which replaces a reset of q0 in a known n-qubit state |ψ⟩ by an n-qubit unitary T_{|ψ⟩→|φ⟩}, where |φ⟩ is \"the state after resetting qubit q0 to 0\" (Eq. 2). Section 3.3 makes the semantics explicit: [[reset_i]]^# s = (1/λ_norm) Σ_{ψ_j^i=0} λ_j |ψ_j⟩. That is projection onto the q0=0 subspace followed by renormalization, i.e., postselection. A physical reset is the completely positive trace-preserving map R_i(ρ) = |0⟩⟨0|_i ⊗ Tr_i(ρ). For an entangled pure input such as (|00⟩+|11⟩)/√2, R_0 leaves q1 maximally mixed, whereas any unitary T on the n qubits maps the pure input to a pure state and leaves q1 pure. No n-qubit unitary can reproduce the mixed reduced state on the remaining qubits. Thus Eq. (2) is not an identity; it changes the state whenever the reset qubit is entangled with the rest. Since QCP has no mixed-state representation, its propagation rule records the postselected pure state rather than the actual mixture, so Algorithm 1 cannot even detect the error. This invalidates the paper's reset-removal claim for entangled states and any reported reset reductions that rely on Theorem 2, including the BV qubit-reuse demonstration if the reset input is entangled. The measurement-reduction part (Theorem 1) is a separate contribution and may remain valid.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the probabilistic circuit model (PCM) to reduce the number of mid-circuit measurements and resets in dynamic quantum circuits. The framework uses quantum constant propagation (QCP) to obtain compile-time knowledge of quantum states, then replaces a mid-circuit measurement on a known state by a probabilistic subcircuit (Theorem 1) and a mid-circuit reset by a unitary state-transformation circuit (Theorem 2). A parameter n_pcm is introduced to trade optimization strength against synthesis cost. The method is evaluated on randomly generated dynamic circuits and on a Bernstein-Vazirani circuit with qubit reuse, where the authors report removing all resets. The paper also improves QCP by separating disentangled qubits and by propagating information through resets.","tokens_in":11939,"tokens_out":7357,"duration_ms":76164,"significance":"If correct, the framework would give a practical compile-time method for reducing dynamic-circuit overhead, complementing existing Qiskit passes, and the public implementation would be a useful resource. However, the central reset-removal claim rests on an incorrect model of reset semantics for entangled states: Section 3.3 defines the abstract reset effect as projection onto the |0> component followed by renormalization, which is postselection, not the CPTP reset performed on hardware. Since Theorem 2 and Algorithm 1 rely on this definition, the reset reductions reported in Section 4 are unsupported for entangled inputs. The measurement-reduction part is more plausible but also needs a precise statement about post-measurement states when the measured qubit is entangled with the rest of the system.","major_comments":[{"comment":"The reset replacement in Theorem 2 is not state-equivalent for entangled input states. The paper's abstract semantics, [[reset_i]]^# s = (1/lambda_norm) sum_{psi_j^i=0} lambda_j |psi_j>, is projection onto the subspace where qubit i is |0>, i.e., postselection. A physical reset is the CPTP map R_i(rho) = |0><0|_i (x) Tr_i(rho). For an entangled pure input such as (|00>+|11>)/sqrt(2), the actual post-reset state of the remaining qubit is maximally mixed, while any n-qubit unitary T maps the pure input to a pure state and leaves the remaining qubit pure. Hence Eq. (2) cannot hold for entangled states, and the QCP propagation rule in Section 3.3 records the postselected pure state rather than the actual mixture. Consequently, Algorithm 1 will replace resets with unitary circuits that change the output state, and the reset-removal results in Section 4, including the Bernstein-Vazirani example in Section 4.4, are unsupported whenever the reset input is entangled.","section":"Section 3.3 and Theorem 2, Eq. (2)"},{"comment":"The measurement replacement in Theorem 1 also needs a careful statement of what is being preserved. Eq. (1) replaces the measurement of an n-qubit state by T_{|psi>->|0...0>} followed by a probabilistic X_s, which produces the computational basis state |s> on all n qubits. For a mid-circuit measurement of a single qubit that is entangled with the rest, the post-measurement state of the unmeasured qubits is a conditional pure state (or a mixture before conditioning on the classical outcome), and it is not reproduced by preparing |s> on the full register. If Theorem 1 is intended only for measurements of all n qubits whose outcomes are the only relevant outputs, then Algorithm 1's application to individual mid-circuit measurements requires an additional argument. As written, the claimed extension of measurement optimization to entangled multi-qubit states is not justified.","section":"Theorem 1 and Algorithm 1, Eq. (1)"},{"comment":"The experimental evaluation reports counts of removed measurements and resets and introduced gates, but it never verifies that the optimized circuits are semantically equivalent to the original circuits, e.g., by comparing output distributions or state fidelities. Given that the reset transformation in Theorem 2 is not state-equivalent for entangled inputs, the reported reductions may partly reflect changes in circuit semantics rather than safe optimizations. A correctness check on the optimized circuits is needed before the empirical claims can be interpreted.","section":"Section 4.3"}],"minor_comments":[{"comment":"There is a typo: 'Dimonstrative example' should be 'Demonstrative example', and the sentence 'applying the PCM-based optimization framework the circuit' is missing 'to'.","section":"Figure 6 caption"},{"comment":"The computation contains a typo: the text writes 'H ⊗ H ∈ C_{|00>}^{|+->}' but then evaluates '(H ⊗ XH)|+->'; the XH should be H.","section":"Example 6"},{"comment":"The notation '1s0 (X0) ⊗ ... ⊗ 1sl−1 (Xl−1)' is confusing; the indicator function should be defined explicitly before use.","section":"Definition 4"},{"comment":"The paper relies on 'runtime equivalence' defined in [6] but does not define it locally; since Theorem 1 and Theorem 2 both use this notion, a brief definition would make the paper more self-contained.","section":"Section 2.3 and Theorem 1"},{"comment":"The text says the data are obtained by evaluating 10 quantum circuits per scale, but the standard-deviation experiment uses 50 circuits; this should be stated explicitly to avoid ambiguity about which plots use which sample size.","section":"Section 4.3"}],"recommendation":"reject","confidential_remarks":"The reset semantics issue is not a minor presentation problem: it invalidates the main claimed contribution of reset removal for entangled states. The measurement-reduction part might be salvageable with a more precise theorem, but as written the central claims are not supported. The paper could be reconsidered after a substantial revision that either restricts the reset transformation to separable states or extends QCP to mixed-state tracking and verifies semantic equivalence empirically."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The reset story is broken, and the measurement story is real. If you take away Theorem 2, you have a competent, well-implemented extension of the authors' own probabilistic circuit model that replaces some mid-circuit measurements by probabilistic static subcircuits; if you keep Theorem 2, the headline reset-removal results are unsupported because reset is modeled as postselection, not as the CPTP map that hardware actually implements. That is the main thing to know about this paper.\n\nWhat's new and good: the n-qubit PCM extension with state synthesis is a natural generalization of the single-qubit pass, and it is actually evaluated on random dynamic circuits with public code. The QCP improvement that separates disentangled qubits from storage groups is a sensible engineering tweak and seems to help tracking. The npcm knob is a clean way to express the trade-off between static synthesis cost and dynamic-operation reduction. Theorem 1, for measurements on known pure states, is plausible: replacing a known-basis measurement by a probabilistic X rotation preserves the distribution on classical wires, and the paper is careful that the measurement outcome is precomputed at compile time.\n\nThe soft spot is not minor. Section 3.3 defines the abstract reset as [[reseti]]^# s = (1/λ_norm) Σ_{ψ_j^i=0} λ_j |ψ_j⟩, which is projection onto the qi=0 subspace followed by renormalization. That is postselection. A physical reset is ρ_i ↦ |0⟩⟨0|_i ⊗ Tr_i(ρ), which for an entangled pure input leaves the other qubits mixed. Theorem 2 instead substitutes a single unitary T|ψ⟩→|φ⟩ on the whole n-qubit register. No unitary on the register can map the pure pre-reset state to a state with a mixed reduced density matrix on the remaining qubits, so Theorem 2 is not state-equivalent whenever the reset qubit is entangled with the rest. The paper's own QCP has no mixed-state representation, so Algorithm 1 cannot detect that it has drifted out of the set of pure states it is tracking. The reset-removal numbers in the evaluation and the claim of removing all resets in the Bernstein-Vazirani qubit-reuse example all inherit this flaw.\n\nCitation-wise, the paper leans heavily on the authors' own prior PCM and QCP papers, but those are the works being extended, so that is appropriate rather than self-promotion. The evaluation methodology is acceptable: random circuits, clear comparisons, and standard-deviation checks at larger scales. There is no parameter fitting to the results; the probabilities come from Born amplitudes, which is a point in the paper's favor.\n\nWho is this for? Compiler people working on dynamic circuits. They would read the measurement section with interest and should be warned that the reset section is semantically wrong. The paper as a whole should not be accepted, but it deserves a serious referee: the measurement-only contribution is publishable as a separate piece, and the reset mistake is exactly what a careful review should catch.","headline":"The measurement-reduction half is a solid, implemented extension of the authors' prior PCM work, but the reset-removal theorem is built on a postselection model of reset that is wrong for entangled states, so the reset claims—including the BV demo—don't hold as stated.","tokens_in":12431,"tokens_out":1880,"would_cite":false,"duration_ms":20256,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that mid-circuit measurements and resets can be replaced at compile time by probabilistic or static subcircuits whenever the measured or reset state is fully known.","keywords":["dynamic circuits","mid-circuit measurements","qubit reset","probabilistic circuit model","quantum constant propagation","circuit optimization","state preparation","Bernstein-Vazirani algorithm"],"falsifier":"Prepare two qubits in the Bell state $(|00\\rangle+|11\\rangle)/\\sqrt{2}$, reset the first qubit, and measure the expectation $\\langle Z\\rangle$ on the second qubit. The paper's projection semantics predicts $\\langle Z\\rangle=1$, while a physical reset yields $\\langle Z\\rangle=0$ because the second qubit becomes maximally mixed.","tokens_in":11406,"feed_emoji":"⚛️","tokens_out":9710,"duration_ms":86234,"temperature":0.7,"pith_summary":"This paper claims that mid-circuit measurements and resets in dynamic quantum circuits can be moved out of the running circuit whenever the state entering the operation is fully known at compile time. It extends the probabilistic circuit model so that a measurement on a known multi-qubit state becomes a probabilistic subcircuit, and a reset becomes a static state-transformation circuit. The payoff is lower runtime overhead: in a Bernstein-Vazirani circuit using qubit reuse, every reset is removed, and across randomly generated dynamic circuits the framework removes over 17% of measurements and resets, compared with under 2% for the reference compiler passes. The parameter $n_{\\mathrm{pcm}}$ lets users trade optimization strength against synthesis time and added gates.","feed_headline":"Compile-time pass deletes all resets in qubit-reuse algorithm","feed_subtitle":"Mid-circuit measurements and resets become probabilistic or static subcircuits, cutting runtime overhead","key_machinery":"The load-bearing object is the extended probabilistic gate $GP[(U_1,p_1),\\dots,(U_k,p_k)]$, which compiles to one of $k$ static circuits according to a compile-time-sampled distribution. Around it the framework combines three components: quantum constant propagation (QCP), enhanced with entanglement-group splitting and reset-propagation rules; the state-transformation circuit $T_{|\\psi\\rangle\\to|\\varphi\\rangle}=SP_{|\\varphi\\rangle}(SP_{|\\psi\\rangle})^{-1}$, built from ancilla-free state preparation; and the parameter $n_{\\mathrm{pcm}}$, which decides whether a fully known state is small enough to synthesize. Theorem 1 uses this machinery to replace known-state measurements with probabilistic $X$-rotations, and Theorem 2 uses it to replace resets with a static transformation. The runtime-equivalence relation from the original probabilistic circuit model carries the correctness argument.","core_discovery":"The central claim is that dynamic-circuit overhead can be converted into static compile-time work. For any mid-circuit measurement whose input $n$-qubit state is fully known and whose support size is at most $n_{\\mathrm{pcm}}$, Theorem 1 shows the measurement can be replaced by a probabilistic subcircuit: synthesize a circuit mapping the known state to $|0\\rangle^{\\otimes n}$, then apply $X_s$ with probability $|\\alpha_s|^2$ and record the sampled string on the original classical wire. For any reset whose input state is fully known, Theorem 2 shows the reset can be replaced by a single static circuit that maps the pre-reset state to the state obtained by setting the target qubit to $|0\\rangle$. This replacement relies on a reset semantics in which the reset projects the state onto the $|0\\rangle$ subspace of that qubit and renormalizes. The evaluation reports that the method removes all resets in the qubit-reuse Bernstein-Vazirani example and outperforms the reference compiler passes on random dynamic circuits.","pith_inferences":["Because the equivalence for resets is stated under a projection semantics, a natural test is to compare the framework's prediction against physical reset on an entangled state; a Bell-state experiment would determine exactly where Theorem 2 applies.","Pre-sampling known measurement outcomes is structurally similar to gate cutting and circuit knitting, so the framework could be combined with those techniques to schedule dynamic algorithms on fewer physical qubits.","The $n_{\\mathrm{pcm}}$ trade-off could be automated by predicting per-site synthesis cost and runtime savings, turning the manual knob into an optimization objective.","The improved QCP with entanglement-group splitting is independent of the reset question and may benefit other static analyses of dynamic circuits."],"forward_implications":["Any dynamic circuit segment whose measured state is fully known can run without real-time feedback for that measurement, since the outcome is sampled at compile time.","Fully known resets can be compiled away entirely, which is what lets qubit-reuse circuits like Bernstein-Vazirani lose all of their resets.","Raising $n_{\\mathrm{pcm}}$ removes additional dynamic operations, at the cost of longer synthesis runs and larger added gate counts, with diminishing returns set by QCP's entanglement-group limit $n_{\\max}$.","Compiling away resets first unblocks downstream structural optimizations, as the transpiled Bernstein-Vazirani circuit becomes much shorter after the framework runs.","The method directly targets the dominant runtime cost of dynamic circuits, which is relevant for algorithms that interleave mid-circuit measurement with classical feed-forward."],"supporting_citations":[{"why":"Supplies the random circuit generator used to build the evaluation dataset of dynamic circuits.","marker":"[3]"},{"why":"Defines the Bernstein-Vazirani algorithm used as the qubit-reuse demonstrative example.","marker":"[4]"},{"why":"Provides the qubit-reuse technique whose resets the framework removes in the demonstration.","marker":"[5]"},{"why":"Introduces the probabilistic circuit model and the original measurement-reduction pass that this work extends.","marker":"[6]"},{"why":"Supplies quantum constant propagation, the static analysis that produces the known input states used by the optimization.","marker":"[8]"},{"why":"Gives asymptotically optimal state-preparation circuit depths underpinning the synthesis bound in Lemma 1.","marker":"[30]"},{"why":"Provides a concrete ancilla-free state-preparation procedure used as the synthesis primitive.","marker":"[34]"}],"fun_headline_variants":["Probabilistic synthesis removes all resets in qubit-reuse algorithm","Compiler pass replaces resets with static circuits, cuts overhead","Optimization framework minimizes mid-circuit measurements and resets","Quantum framework converts known-state resets into static gates","Dynamic circuit optimization: resets become static subcircuits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The framework assumes that resetting a qubit is equivalent to projecting the whole state onto the $|0\\rangle$ subspace of that qubit and renormalizing, so a single static subcircuit can reproduce the post-reset state; a physical reset of an entangled qubit leaves the other qubits mixed, which no unitary subcircuit can match.","fun_headline_variants_meta":{"raw":{"variants":["Probabilistic synthesis removes all resets in qubit-reuse algorithm","Compiler pass replaces resets with static circuits, cuts overhead","Optimization framework minimizes mid-circuit measurements and resets","Quantum framework converts known-state resets into static gates","Dynamic circuit optimization: resets become static subcircuits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000785,"raw_usage":{"total_tokens":3445,"prompt_tokens":903,"completion_tokens":2542,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":2460}},"tokens_in":519,"tokens_out":2542,"duration_ms":18157,"temperature":1.0,"reasoning_tokens":2460,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:01:20.171658+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Prepare two qubits in the Bell state $(|00\\rangle+|11\\rangle)/\\sqrt{2}$, reset the first qubit, and measure the expectation $\\langle Z\\rangle$ on the second qubit. The paper's projection semantics predicts $\\langle Z\\rangle=1$, while a physical reset yields $\\langle Z\\rangle=0$ because the second qubit becomes maximally mixed.","supporting_citations":[{"cited_title":"https://docs.quantum.ibm.com/api/qiskit/0.19/qiskit.circuit.random.random_circ","cited_arxiv_id":null,"evidence_quote":"Supplies the random circuit generator used to build the evaluation dataset of dynamic circuits."},{"cited_title":"Physical Review A 109(4), 042401 (2024)","cited_arxiv_id":null,"evidence_quote":"Provides a concrete ancilla-free state-preparation procedure used as the synthesis primitive."}],"review_version":1}