{"id":"45caabd3-a2c5-4544-830a-ff868e92cef0","arxiv_id":"2502.01008","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For a fully connected ferromagnetic Ising model, fixing spins improves quantum annealing only when the rescaling parameter is small; at large rescaling it hurts.","lead":"Fixing spins before running a quantum annealer can shrink the problem, but the benefit depends on how much the energies are rescaled to fit the hardware. On a simple ferromagnetic test model, spin fixing helps when the rescaling parameter is small and hurts when it is large.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim is conditional on a known ground state; the paper's own Section 2 admits p_err is only computable for trivial ferromagnetic case, so the general conclusion overreaches.","rationale":"The reader's weakest assumption correctly identifies that p_err is only computable for the trivial ferromagnetic ground state and that the paper's method characterization is circular for real problems. The strongest claim—that fixing spins helps at small r and hurts at large r—is internally consistent within the homogeneous fully connected model, and the finite-size numerics plus thermodynamic-limit bosonization agree without fitted parameters. However, the abstract and conclusion state that 'the fixing spins method enhances quantum annealing performance' as a general statement, while the only tested case has a known ground state and p_err is an input. This is not an internal inconsistency but a scope limitation that invalidates the generalized claim. The proposed test would directly address whether the single-parameter p_err reduction survives when fixing decisions come from a realistic preprocessing method. Until such evidence exists, the verdict should remain CONDITIONAL, not ACCEPT or REJECT: the paper's contribution is a well-analyzed special-case trade-off, which is valuable, but the advertised conclusion about fixing spins as a general size-reduction method is not yet established.","tokens_in":150,"tokens_out":817,"duration_ms":23427,"concrete_test":"Run the same finite-size exact-diagonalization and hardware analysis on a random fully connected ferromagnetic Ising model with random local fields (or a Sherrington–Kirkpatrick model) where fixing spins are chosen by a realistic sample-persistence preprocessing instead of by the known ground state. Compare the resulting hardware minimum energies and minimum gaps, as functions of r and n, against the curves predicted by Eq. (14) using the measured p_err. If the performance deviates beyond the scatter expected from 100 runs, the single-scalar p_err reduction fails for preprocessing-generated fixes and the general claim is unsupported. Additionally, report the distribution of p_err from the preprocessing method to show whether low-error fixing is actually achievable in the tested regime.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central claim—that fixing spins improves quantum annealing at small r and degrades it at large r—rests on sweeping the fixing error p_err into a single scalar field shift h' = h + J(N−n)(1−2p_err), Eq. (14). This reduction is only valid for the homogeneous fully connected ferromagnetic model, where the ground state is the known all-up state and p_err is an input parameter, not an output of a preprocessing method. Section 2 explicitly states: 'the error probability is easy to calculate since a homogeneous fully connected ferromagnetic Ising model, which has the trivial ground state, is used.' For real optimization problems, the ground state is unknown, so p_err cannot be controlled or even measured a priori. A real fixing-spins method (e.g., sample persistence) will produce spin directions with a distribution of errors that is correlated with the problem structure and with the local fields; these correlations are not captured by a single uniform scalar shift. The paper's own Hardware results in Fig. 2(d)–(f) show that large p_err leads to chain breaks and high energies, but the analysis treats p_err as an exogenous parameter. The conclusion that one must use 'a fixing spins method with a low error probability' is therefore a restatement of the assumption rather than a derived result. The paper does not demonstrate that any practical preprocessing method can achieve the low p_err needed for the claimed benefit, nor that the interplay with rescaling remains the same when errors are spatially correlated. This is the load-bearing gap: the demonstrated trade-off is for an idealized, ground-state-informed fixing procedure, while the abstract and conclusion generalize to size-reduction methods for real combinatorial problems.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the interplay between fixing spins (a size-reduction method) and energy rescaling in quantum annealing, using the homogeneous fully connected ferromagnetic Ising model as a testbed. It introduces an error probability p_err for mis-fixed spins, derives the effective local field h' after fixing spins (Eq. 14), and analyzes the minimum energy gap via exact diagonalization for N=160 and via a Holstein-Primakoff/Bogoliubov thermodynamic-limit formula. Hardware experiments on D-Wave Advantage measure the minimum energy over 100 runs as a function of the reduced size n and the rescaling parameter r. The central claim is that fixing spins improves quantum annealing performance at small r (wide energy range) and degrades it at large r, with an optimal n that depends on r and p_err.","tokens_in":19517,"tokens_out":6721,"duration_ms":64521,"significance":"If the results hold, they provide a concrete and largely parameter-free analysis of a nontrivial trade-off: fixing spins increases the energy gap through reduced system size but simultaneously modifies the local fields, and the benefit depends on the rescaling parameter r. The thermodynamic-limit gap formula and the explicit dependence of h' on p_err are useful benchmarks. The paper is transparent about its limitation to a trivial-ground-state model, and the analytic derivations are reproducible. The hardware experiment is illustrative, but it is not statistically characterized and should not be the basis for quantitative claims.","major_comments":[{"comment":"The experimental claim that fixing spins enhances the quantum annealer's performance is supported only by the minimum energy among 100 stochastic runs, with no error bars, standard deviation, or repeated-measurement statistics. Since D-Wave outputs are stochastic, the differences among cells in Fig. 2 could be noise; please provide a statistical characterization (e.g., error bars, median, or distribution of chain breaks) to justify the claim.","section":"3.1, Fig. 2"},{"comment":"The error probability p_err is an exogenous parameter that is only well-defined because the ground state is known (all-up state). The paper acknowledges this in Sec. 2, but the Conclusion generalizes to 'a fixing spins method with a low error probability is essential' and proposes hardware design changes. For real optimization problems, p_err is not known a priori and errors are not uniform; Eq. (14) does not capture correlations between fixed-spin errors and local fields. Please scope the conclusions to the homogeneous ferromagnetic model, or add a discussion of what changes when the ground state is unknown.","section":"2 and Conclusion"},{"comment":"The statement 'Fixing spins has a positive effect at small r but a negative effect at large r' (Sec. 3.2) is a useful summary but is too coarse for intermediate r: Fig. 4 shows non-monotonic behavior with an optimal n for r around 2–3, and even at p_err = 0.5 the gap varies with n because the system size changes. Please characterize the optimal n as a function of r and p_err, or at least qualify the claim to reflect the non-monotonic structure.","section":"3.2, Fig. 4"}],"minor_comments":[{"comment":"Equation (15) is typeset with a brace but no cases; the expression runs across lines without a clear branch structure. Please format it as a single display equation or a proper cases environment.","section":"Eq. (15)"},{"comment":"The phrase 'shows the magnetic process of Δmin' should be 'shows the s-dependence of Δmin' or similar.","section":"3.2, Fig. 5 caption"},{"comment":"The claim that 'the properties of the energy gap with fixing spins and rescaling are extensive with respect to size' is not established; the thermodynamic limit is a specific N→∞ limit, not evidence of extensivity. Please rephrase.","section":"3.2"},{"comment":"Equation (12) uses subscript i without distinguishing the reduced-system index i′; this is a minor notation inconsistency.","section":"2, Eq. (12)"},{"comment":"The caption says 'The ground state energy is 0.597' but does not specify the units or the parameter values used to compute this; please clarify.","section":"Fig. 2 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of JPSJ and the theoretical analysis is sound for the toy model. The main risks are the unquantified hardware experiment and the tendency to overgeneralize conclusions beyond the homogeneous ferromagnetic model. The experimental section would benefit from a statistical review. I recommend major revision rather than rejection because the core analytic results are defensible and the weaknesses are local and fixable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing to know: this paper delivers a clean, narrow result on the interplay between spin-fixing size reduction and energy rescaling in a quantum annealer, for a homogeneous fully connected ferromagnetic Ising model. The core finding --- fixing spins helps at small r (wide energy range) and hurts at large r --- is supported by exact diagonalization and a thermodynamic-limit bosonization that agree with no fitted parameters. The hardware data on a D-Wave device is consistent with the trend, though it is presented as the minimum of 100 runs without error bars, so treat the experimental part as qualitative.\n\nWhat is new: prior work treated fixing spins without rescaling (Ref. 21) and rescaling without fixing (Refs. 35, 36). Combining them exposes a trade-off that matters for parameter tuning: the rescaling parameter controls whether size reduction helps or hurts the gap. Eq. (14) shows the effect is mediated by the single scalar h' = h + J(N-n)(1-2p_err). The thermodynamic-limit calculation is a standard but competently executed Holstein-Primakoff plus Bogoliubov derivation.\n\nSoft spots, in order of importance. First, the central quantity p_err is an input, not an output. The paper says openly that it is easy to calculate because the ground state is trivially known; for real optimization problems you do not know the ground state, and a real preprocessing method produces correlated errors that Eq. (14) cannot capture. So the demonstrated trade-off is for an idealized, ground-state-informed fixing procedure. That is not a fatal flaw for the model studied, but it means the abstract and conclusion overreach when they generalize to “the fixing spins method” without qualification. Second, the hardware experiment uses only the minimum of 100 runs and no variance estimate; chain-break effects are discussed qualitatively but not quantified. Third, the claim in Sec. 3.2 that finite-size results extend to larger sizes rests on the thermodynamic-limit comparison; reasonable, but not a proof of extensivity for the gap.\n\nNone of these undercut the core finite-size result for this model. The paper is honestly written, acknowledges the simplest setup, and points to future work on random and constrained models. I would send it to peer review; a referee should push for a scope-limited abstract and a statistical treatment of the hardware runs. A conditional accept is appropriate, not a rejection.","headline":"A narrow but genuine result: fixing spins helps or hurts quantum annealing depending on the energy rescaling parameter, demonstrated cleanly for a homogeneous ferromagnetic model, though the practical generalization is limited by the known-ground-state assumption.","tokens_in":20173,"tokens_out":2410,"would_cite":false,"duration_ms":22687,"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":"Fixing spins helps or hurts a quantum annealer depending on the energy rescaling.","keywords":["quantum annealing","fixing spins","energy rescaling","minimum energy gap","fully connected ferromagnetic Ising model","minor embedding","qubit chains","thermodynamic limit"],"falsifier":"Reproduce the exact diagonalization of the $N=160$ homogeneous fully connected ferromagnetic model with $p_{\\mathrm{err}}=0$: the claimed crossover requires that at $r=0.25$ the minimum gap is larger when fewer spins remain and falls as $n$ grows to 160, while at $r=10$ the gap rises with $n$; if either ordering fails, the central crossover claim is wrong.","tokens_in":19029,"feed_emoji":"🧲","tokens_out":12256,"duration_ms":107231,"temperature":0.7,"pith_summary":"This paper asks whether the fixing-spins size-reduction method, which deletes a subset of variables before annealing, still helps when the hardware forces the problem's energy scale to be rescaled. Working with a homogeneous fully connected ferromagnetic Ising model (all pairs of spins share the same attractive interaction, plus a small magnetic field to remove degeneracy), the paper finds the answer is conditional on the rescaling parameter $r$. At small $r$, fixing spins improves both the lowest energy reached on the annealer and the minimum energy gap; at large $r$, fixing spins shrinks the gap and harms the search. The connecting mechanism is that fixed spins shift the local field to $h' = h + J(N-n)(1-2p_{\\mathrm{err}})$, which changes the energy range and hence the rescaling that the hardware applies. The result is supported by experiments on a quantum annealer, by exact diagonalization of the gap, and by a thermodynamic-limit calculation.","feed_headline":"Fixing spins helps or hurts annealing depending on rescaling","feed_subtitle":"In a ferromagnetic model, small rescaling makes spin fixing a gain; large rescaling makes it a loss.","key_machinery":"The carrying object is the reduced local field $h' = h + J(N-n)(1-2p_{\\mathrm{err}})$, a scalar that converts the whole fixing-spins procedure into one shift of the magnetic field. Rescaling by $r$ turns the problem Hamiltonian into $H''_p = H'_p/r$, so $h'$ sets both the energy range available on the hardware and the effective annealing schedule $B'(s) = B(s)/r$. In the thermodynamic limit the model is reduced by a Holstein-Primakoff transformation (a standard mapping of spin operators to creation and annihilation operators) and then diagonalized by a Bogoliubov transformation, yielding a harmonic oscillator whose frequency, minimized over $s$, is the gap $\\Delta_{\\min}$. A second mechanism enters on real hardware: qubit chains from minor embedding break when $h'$ is small relative to the fixed chain strength, which happens for large $n$ and large $p_{\\mathrm{err}}$ or small $r$.","core_discovery":"The central claim is that the benefit of fixing spins cannot be judged on its own: it is controlled by the rescaling parameter $r$. For the homogeneous fully connected ferromagnetic model with $N$ spins, fixing $n$ of them with error probability $p_{\\mathrm{err}}$ yields an equivalent reduced model with unchanged coupling $J$ and modified local field $h' = h + J(N-n)(1-2p_{\\mathrm{err}})$. Since the hardware rescales the problem parameters by $r$, this shift in $h'$ changes the actual Hamiltonian that is annealed. The paper reports a crossover: with a wide energy range (small $r$), fixing spins lowers the minimum energy and enlarges the minimum gap, while with a narrow range (large $r$) the gap-shrinking effect of rescaling dominates and fixing spins degrades performance. The experiments and exact-diagonalization results agree on this crossover, and the thermodynamic-limit analysis shows the gap increases with $h'$ and has an optimal rescaling value for a fixed parameter range.","pith_inferences":["For inhomogeneous or random Ising problems the scalar $h'$ formula no longer holds, so the $r$-dependent crossover found here may be weaker or reversed; sweeping $r$ while fixing spins on such models would show whether the trade-off is generic.","The same argument suggests a practical protocol: choose the rescaling parameter and the number of fixed spins jointly, since optimizing either alone misses the region where both are favorable.","If a future annealer reduced chain length by adding couplers per qubit, the chain-break penalty at small $r$ would shrink, possibly making aggressive spin fixing useful over a wider rescaling range.","In real optimization the ground state is not known, so $p_{\\mathrm{err}}$ cannot be fixed in advance; whether classical preprocessing can keep the effective error low enough to sit in the beneficial small-$r$ regime remains an open question."],"forward_implications":["When the energy scale must be rescaled, the optimal number of fixed spins depends on $r$; for instance, at $r=2.5$ the gap is largest at $n=140$ rather than at the most aggressive reduction.","Fixing spins in the correct direction is doubly useful on real hardware: it both lowers the energy and strengthens $h'$, which protects the embedded qubit chains from breaking.","At error probability $p_{\\mathrm{err}}=0.5$, $h'$ returns to the original field $h$, so the energy range is unchanged and fixing spins loses its rescaling-related advantage.","For small $r$ the trade-off favors fixing spins; for large $r$ rescaling penalizes it, so the number of spins to fix and the amount of rescaling should be treated as one tuning decision."],"supporting_citations":[{"why":"It defines the fixing-spins method via the error probability $p_{\\mathrm{err}}$ and establishes the no-rescaling baseline that fixing spins enlarges the gap.","marker":"21)"},{"why":"It defines minor embedding, the mapping that creates the qubit chains whose breaking is central to the hardware results.","marker":"28)"},{"why":"It extends minor embedding to hardware graphs, grounding the chain-length and chain-strength discussion.","marker":"29)"},{"why":"It supplies the adiabatic theorem linking annealing time to the minimum gap, making $\\Delta_{\\min}$ the performance metric.","marker":"37)"},{"why":"It supplies the hardware parameter settings, including auto-scaling and the maximum chain strength, used in the experiment.","marker":"40)"},{"why":"It provides the analytic thermodynamic-limit method that yields the harmonic-oscillator gap formula the paper extends to fixing spins and rescaling.","marker":"42)"},{"why":"It gives the Holstein-Primakoff transformation used in Appendix A to bosonize the total spin Hamiltonian.","marker":"43)"},{"why":"It gives the Bogoliubov transformation that diagonalizes the bosonized Hamiltonian and produces the excitation gap.","marker":"44)"}],"fun_headline_variants":["Fixing spins: good or bad? Depends on rescaling","Quantum annealing: spin fixing's fate hinges on rescaling","Spin fixing effect in annealing tied to energy rescaling","Rescaling decides if fixing spins aids annealing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument presupposes that the ground state is known well enough to define the error probability $p_{\\mathrm{err}}$, which is only true for the trivial all-up state of this ferromagnetic model; for real optimization problems the ground state is unknown, so $p_{\\mathrm{err}}$ cannot be controlled before annealing and wrongly fixed spins may have effects beyond the single scalar $h'$.","fun_headline_variants_meta":{"raw":{"variants":["Fixing spins: good or bad? Depends on rescaling","Quantum annealing: spin fixing's fate hinges on rescaling","Spin fixing effect in annealing tied to energy rescaling","Rescaling decides if fixing spins aids annealing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000272,"raw_usage":{"total_tokens":1622,"prompt_tokens":928,"completion_tokens":694,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":544,"completion_tokens_details":{"reasoning_tokens":628}},"tokens_in":544,"tokens_out":694,"duration_ms":6165,"temperature":1.0,"reasoning_tokens":628,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T16:53:54.359638+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Reproduce the exact diagonalization of the $N=160$ homogeneous fully connected ferromagnetic model with $p_{\\mathrm{err}}=0$: the claimed crossover requires that at $r=0.25$ the minimum gap is larger when fewer spins remain and falls as $n$ grows to 160, while at $r=10$ the gap rises with $n$; if either ordering fails, the central crossover claim is wrong.","supporting_citations":[],"review_version":1}