{"id":"27376168-4067-49ba-974b-15542e1ef021","arxiv_id":"2411.16009","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Spontaneous polarization symmetry breaking in a driven fiber Kerr resonator is used to encode Ising spins, demonstrated with up to 100 spins, intensity readout, and hour-long stable operation.","lead":"This paper demonstrates a new kind of optical Ising machine that represents binary spins as two polarization states of light in a fiber resonator, read out by simple intensity measurements. The device ran continuously for over an hour with up to 100 coupled spins, and its solution time grows roughly like exp(sqrt(N)).","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Bias-free spin statistics under feedback are not directly demonstrated; a binomial ground-state test would settle whether the topological protection holds in the coupled machine.","rationale":"Reader's weakest assumption is topological protection; I agree it is the most load-bearing. The paper's central novelty is a bias-free Ising emulation, and the only in-paper support for 'no unwanted biases' is a self-cited theory and agreement with an idealised simulation. That agreement is necessary but not sufficient: the simulation omits the very imperfections the topological argument is supposed to defeat. The proposed test uses only data already collected (or a small deliberate perturbation experiment) and directly interrogates the symmetry of the physical spin statistics, which is the quantity the central claim cares about. I do not request rejection; the demonstration is credible and the reader's CONDITIONAL verdict remains appropriate, so verdict_should_be is UNCHANGED. The secondary issue of undisclosed parameters (α, δ0, θ, χ, g) reinforces the need for this test, since it prevents independent verification of the simulation-based validation.","tokens_in":10966,"tokens_out":5575,"duration_ms":59104,"concrete_test":"Using the existing 1,500-trial 64-spin dataset, compute the per-spin magnetisation M_k = (1/T)Σ_t σ_k^t and the frequency f_+ of the two degenerate AF ground states (alternating patterns). Under topological protection, f_+ should be 0.5 within binomial error (sqrt(0.25/1500)≈1.3%) and every |M_k| should be within ~2σ of 0. Repeat the measurement after deliberately detuning PC2 away from π by 5°–10° and after misaligning the input drive polarization; if |M_k| develops a systematic offset or f_+ departs from 0.5, the protection does not survive the feedback-coupled regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Refs 28/29 establish bias-free spontaneous polarization symmetry breaking for an isolated Kerr resonator. In the Ising machine, each spin is perturbed by a feedback term phi_i^(m+1) = g Σ_j J_ij |E_{+,j}^{(m)}|² (Methods, Eq. 4) and by the deliberate E2 driving component; this is precisely the class of symmetry-breaking perturbation that the topological protection must suppress, but no direct bias measurement is reported. The comparison in Fig. 4 is against the authors' own Ikeda model (Eqs. 2–5), which contains no imperfect π phase shift, no polarization misalignment, and no polarization-dependent loss, and whose parameters (α, δ0, θ, χ, g, noise level) are not stated. Agreement with that idealised model cannot certify that the physical spins are unbiased. If the protection fails in the coupled, pulsed, feedback-driven regime, the measured energies would be those of a biased Hamiltonian, undermining the claim of an Ising machine without post-selection. The stability over one hour, while impressive, only shows the bias is stable, not that it is zero.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental coherent Ising machine based on spontaneous polarization symmetry breaking in a coherently driven fibre Kerr resonator. Artificial spins are encoded in the two phases of a parametrically generated orthogonal polarization mode, and the spin state is read out by measuring the intensity of a hybridized polarization mode rather than by phase-sensitive homodyne detection. The authors demonstrate a one-dimensional antiferromagnetic chain of up to 100 time-multiplexed spins, compare 1,500-trial energy statistics with a numerical Ikeda model, report stable operation over one hour without manual adjustment or post-selection, and interpret their time-to-solution data as consistent with exp(sqrt(N)) scaling. The central experimental platform is novel and the stability data are genuinely impressive, but two load-bearing claims — bias-free operation under feedback and the exp(sqrt(N)) scaling — require stronger support than the manuscript currently provides.","tokens_in":11184,"tokens_out":6684,"duration_ms":64157,"significance":"If the central claims hold, this is a valuable addition to the coherent Ising machine landscape: intensity-only readout, an all-fibre telecom-wavelength implementation, and long-term stability without post-selection would address well-known limitations of DOPO-based machines. The experiment is internally consistent: the 1,500-trial energy distribution, the hour-long stability measurement, the explicit statement that no trials were rejected, and the comparison with an explicit numerical model are all strengths. The main uncertainties concern whether the topological symmetry protection demonstrated in an isolated resonator in ref. 28 carries over to the coupled, feedback-driven, pulsed regime, and whether the scaling claim is an extrapolation from an in-sample fit rather than a demonstrated property. These issues are correctable with additional experiments and reanalysis, so they do not undermine the value of the platform itself.","major_comments":[{"comment":"The exp(sqrt(N)) scaling claim is not an independent prediction: the curve is fitted to the N=10 to 56 data and then compared to the entire dataset, so the reported R^2 values are in-sample for the majority of the points. The statement in the text that the exp(sqrt(N)) curve \"consistently serves as an upper bound for the data\" is weak evidence, since an upper bound does not establish the functional form. Please report out-of-sample errors on the N=64, 80, and 100 points, give parameter uncertainties for the fits, and compare against a more general form such as exp(a N^b) with b free, or explicitly reframe the scaling claim as a hypothesis for future work rather than a demonstrated property of the machine.","section":"Results, Fig. 5; Methods, \"Determination of time-to-solutions\""},{"comment":"No direct bias measurement is reported in the coupled, feedback-driven regime. The feedback term phi_i^(m+1) = g sum_j J_ij |E_{+,j}^{(m)}|^2 (Methods Eq. 4), together with the deliberate E2 driving component, is exactly the class of symmetry-breaking perturbation that the topological protection of ref. 28 must suppress, but that protection was derived for an isolated resonator, not for the coupled machine. Agreement with the idealised Ikeda map (Eqs. 2-5) cannot certify that the physical spins are unbiased, because that model contains no imperfect pi phase shift, no polarization misalignment, and no polarization-dependent loss. A direct test would be to compare, over many trials, the frequencies of the two degenerate ground-state spin configurations of the antiferromagnetic chain, or to measure per-spin +1/-1 counts with the coupling disabled. Without such a test, the claims that the spins evolve \"without unwanted biases\" and that no post-selection is needed are not fully supported.","section":"Methods, Eq. (4); Results, Fig. 4"},{"comment":"The simulated distributions used for comparison in Figs. 3c and 4a depend on parameters that are not stated: alpha, delta_0, theta, chi, g, and the noise amplitude. The text says only that \"environmental noise is represented by adding weak uncorrelated white noise\" without giving its strength or calibration. This makes it impossible to assess whether the \"excellent agreement\" in Fig. 4a is obtained with physical parameter values or with parameter tuning. Please provide a table of all model parameters and, where possible, measured values with uncertainties; also specify how the noise amplitude relates to the experimental conditions.","section":"Methods, \"Numerical model\""}],"minor_comments":[{"comment":"The caption says \"The bars in black correspond to one standard deviation,\" but it is not clear whether this is the standard deviation across batches of 1,500 trials or across individual trials; please specify.","section":"Fig. 4b"},{"comment":"The Methods state that each time-to-solution point is derived from 1,500-4,500 runs, but the exact number of runs and the criterion for selecting the optimal annealing time (beyond locating the minimum of T_s) should be stated explicitly, including any smoothing or interpolation used.","section":"Fig. 5a"},{"comment":"The notation in the nonlinear coupling terms is easy to misread: please rewrite Eq. (5) with explicit mode indices, e.g., E_l^* E_{3-l}^2, and define the coefficients B and C in the text.","section":"Eq. (5)"},{"comment":"The inset shows \"defects\" in the final spin string, but the figure caption does not define how a defect is identified; please state the criterion used.","section":"Fig. 3c"},{"comment":"The claim that all-to-all coupling can be \"readily achieved\" using FPGA-based methods is plausible but is not demonstrated here; since the present experiment is limited to a 1D chain, this sentence should be phrased as a future direction rather than as an established capability.","section":"Discussion"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially a good fit for the journal, but the current version overstates two load-bearing conclusions: the exp(sqrt(N)) scaling is an in-sample fit extrapolation, and the bias-free claim rests on an unverified transfer of topological protection from the isolated resonator to the coupled feedback-driven machine. I would encourage the editor to require a direct symmetry/bias test and an out-of-sample scaling analysis, or a clear softening of both claims, before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read Quinn et al. on the polarization coherent Ising machine. The genuinely new thing is real: spins encoded in the polarization symmetry-breaking of a coherently driven fibre Kerr resonator, read out by intensity alone, with no post-selection. The hour-scale continuous operation from an all-telecom-fibre setup is a concrete practical gain over DOPO CIMs, and the 1,500-trial energy distribution matching their Ikeda model is internally consistent evidence the machine does what it claims.\n\nTwo soft spots, in proportion. First, the exp(sqrt(N)) scaling claim. It is a two-parameter fit to the N = 10–56 subset, with R-squared evaluated on the full set. Credit where due: they disclose the procedure, and the later data points sitting under the curve is a decent out-of-sample check. But over N ≤ 100, exp(sqrt(N)) and exp(N) are not far enough apart that this dataset can sharply separate them. Reading this as “consistent with” is fair; the paper’s “strong evidence” wording overshoots.\n\nSecond, and more important for the machine’s credibility, the bias-free claim. The topological protection they lean on (ref. 28) was established for an isolated resonator. Here every spin receives a feedback perturbation phi_i = g sum J |E_+|^2 and a deliberate E2 driving component—precisely the class of asymmetry the protection is meant to suppress. The evidence offered is agreement between the measured energy distribution and their own idealized model, which has no imperfect pi shift, no polarization misalignment, no polarization-dependent loss, and whose parameter values (alpha, delta0, chi, g, noise amplitude) are never stated. That agreement is suggestive but indirect, and with undisclosed free parameters it cannot carry the full weight. A cheap, decisive addition: report how often each of the two degenerate ground states of the 1D antiferromagnetic chain occurs, or the per-spin marginal rates, and run a binomial test. The hour-long stability tells you the bias is stable, not that it is zero.\n\nMinor point: a 1D AF chain is a weak benchmark, and the authors are appropriately modest about that.\n\nOverall, a novel platform, honestly reported, with fixable weaknesses. It deserves a serious referee; the revision should ask for the model parameters, the direct bias test, and tempered scaling language. I would cite it and bring it to the reading group.","headline":"Polarization-SSB Kerr-resonator Ising machine: genuinely new platform with a solid hour-scale stability demo, but the exp(sqrt(N)) scaling claim is a self-fit and bias-freeness rests on an idealized, under-specified model comparison.","tokens_in":11729,"tokens_out":5303,"would_cite":true,"duration_ms":50608,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Spontaneous polarization symmetry breaking in a fibre Kerr resonator can serve as the spin mechanism of an optical Ising machine, with intensity-only readout and hour-long stable operation.","keywords":["coherent Ising machine","polarization symmetry breaking","Kerr nonlinear resonator","optical fibre ring resonator","time-multiplexed spins","Ising spin chain","combinatorial optimization","symmetry protection"],"falsifier":"With coupling disabled, run many single-spin trials while deliberately misaligning the input polarization or detuning PC2 from its π setting; if the probability of +1 versus −1 departs from 50/50, or if the final-energy distribution of 64-spin chains shifts measurably, the topological symmetry protection on which the bias-free claim rests has failed.","tokens_in":26,"feed_emoji":"💡","tokens_out":7857,"duration_ms":131541,"temperature":0.7,"pith_summary":"This paper claims a new class of optical Ising machine: instead of encoding binary spins in the phase of degenerate optical parametric oscillators, it encodes them in the polarization of pulses circulating in a coherently driven optical fibre Kerr resonator. Spontaneous polarization symmetry breaking gives each pulse two equally probable alternating intensity patterns, which are read out as +1 or −1 by a simple intensity measurement. A localized birefringent defect inside the resonator is claimed to protect the system's symmetry against practical asymmetries, so spins stay unbiased without post-selection. Proof-of-concept runs on chains of up to 100 anti-ferromagnetically coupled spins ran continuously for over an hour without manual adjustment, and the scaling of time-to-solution is consistent with exp(√N). If correct, this points toward Ising machines that avoid phase stabilization and are built entirely from telecom fibre components.","feed_headline":"Polarization-encoded spins run an Ising machine for an hour straight","feed_subtitle":"Intensity-only readout replaces phase-sensitive detection; 100-spin trials run for over an hour without reset.","key_machinery":"The central machinery is a coherently driven passive Kerr fibre ring resonator with two orthogonal polarization modes, plus a localized birefringent defect (polarization controller PC2) that imparts a π phase shift every round trip. Near the tilted Kerr resonance, the undriven mode E2 is parametrically generated with two bistable phases φ0 and φ0+π; the defect swaps these phases every round trip, so the hybrid modes E±=(E1±iE2)/√2 alternate between high and low intensity. The sequence (high, low, high, low, ...) or (low, high, low, high, ...) defines spin +1 or −1, and a single photodiode measurement tells which spin was realised. Coupling between spins is implemented by measuring the instantaneous |E+|^2 of each pulse and feeding back a phase-modulated weak drive along E2, giving the Ikeda-map update $φ_i^{{(m+1)}}$ = g Σ_j J_ij |E_{+,j}^{(m)}|^2. The π-shift defect is what protects this scheme from asymmetries, converting a fragile pitchfork bifurcation into a symmetry-protected one.","core_discovery":"The paper reports an experimental optical Ising machine in which artificial spins are carried by the polarization state of pulses in a coherently driven fibre Kerr resonator rather than by the optical phase of parametric oscillators. Because the two spin values map to two distinct alternating intensity patterns in hybrid polarization modes, readout is a direct intensity measurement, eliminating the homodyne phase detection and phase stabilization used in DOPO-based coherent Ising machines. Operating in the symmetry-protected regime introduced in the paper's reference [28], the machine ran continuously on one-dimensional anti-ferromagnetic spin chains of up to 100 spins for over an hour with no reset and no post-selection, produced final-energy distributions matching numerical simulations, and showed a time-to-solution scaling consistent with exp(√N).","pith_inferences":["A testable extension not pursued in the paper would be to replace the fixed optical delay lines with an FPGA-based feedback loop to implement all-to-all coupling, which the paper says is readily achievable.","The reported hour-long stability does not bound longer-term drift; a natural stress test is to run the machine for days while monitoring the spin-state histogram for the emergence of bias.","Because spin selection is seeded by noise at a pitchfork bifurcation, the same polarization-encoding scheme could double as a high-speed physical random number generator; the paper notes the connection to earlier work but does not claim this as a result of the present study."],"forward_implications":["If the central claim is correct, Ising machines can be built without homodyne phase detection; readout becomes a photodiode intensity measurement on one hybrid polarization mode.","The same telecom fibre components at 1550 nm can support continuous, unattended operation for at least an hour with no post-selection, making long experiments and repeated trials routine.","Because each pulse is an independent spin and pulses are time-multiplexed, the number of spins is adjustable with a pulse picker; chains up to 100 spins were demonstrated.","Time-to-solution scales as exp(√N) for one-dimensional chains with nearest-neighbour coupling, suggesting the machine remains competitive as problem size grows, at least within the demonstrated range of N ≤ 100.","All-to-all coupling should be reachable by replacing the fixed delay-line feedback with established FPGA-based measurement-feedback techniques; the paper states this is readily achievable but does not demonstrate it."],"supporting_citations":[{"why":"Supplies the nonlinear topological symmetry protection mechanism (localized birefringent defect) on which the claim of bias-free spins rests.","marker":"[28]"},{"why":"Establishes that spontaneous symmetry breaking in a Kerr resonator yields unbiased random binary states read out via stroboscopic intensity measurements, the basis for the spin encoding.","marker":"[29]"},{"why":"Original coherent Ising machine proposal based on degenerate optical parametric oscillators; defines the Ising Hamiltonian and phase encoding that the paper replaces with polarization.","marker":"[6]"},{"why":"Demonstrates time-multiplexed optical parametric oscillator networks as coherent Ising machines, the architecture the polarization scheme extends.","marker":"[7]"},{"why":"Large-scale coherent Ising machine showing the phase-stabilization and post-selection challenges the paper aims to avoid.","marker":"[8]"},{"why":"Fully programmable 100-spin coherent Ising machine with all-to-all connections; provides the FPGA feedback methods the paper cites for future all-to-all coupling.","marker":"[9]"},{"why":"Establishes polarization multistability and instability in nonlinear dispersive ring cavities, the physical effect behind the polarization spins.","marker":"[25]"},{"why":"Provides the symmetry-breaking analysis for driven Kerr resonators used to interpret the polarization spontaneous symmetry breaking bifurcation.","marker":"[26]"},{"why":"Supplies the performance comparison against classical neural networks and the scaling benchmark that the paper's time-to-solution data are compared with.","marker":"[33]"}],"fun_headline_variants":["Polarization-encoded Ising machine runs an hour without reset","Intensity readout gives hour-long stable optical Ising machine","Kerr resonator Ising machine: 100 spins, no reset for an hour","All-fibre polarization Ising machine runs continuously for an hour","Polarization spins make hour-long Ising machine with simple readout"],"cache_read_input_tokens":13952,"weakest_assumption_plain":"The machine is bias-free only if the single birefringent defect inside the resonator really cancels every asymmetry, including an imperfect π phase shift and a misaligned driving polarization, as claimed in the paper's earlier theoretical work; if that protection fails, the spins will prefer one state and the measured energy statistics will not represent the intended Ising problem.","fun_headline_variants_meta":{"raw":{"variants":["Polarization-encoded Ising machine runs an hour without reset","Intensity readout gives hour-long stable optical Ising machine","Kerr resonator Ising machine: 100 spins, no reset for an hour","All-fibre polarization Ising machine runs continuously for an hour","Polarization spins make hour-long Ising machine with simple readout"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001315,"raw_usage":{"total_tokens":5349,"prompt_tokens":929,"completion_tokens":4420,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":545,"completion_tokens_details":{"reasoning_tokens":4327}},"tokens_in":545,"tokens_out":4420,"duration_ms":24485,"temperature":1.0,"reasoning_tokens":4327,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:37:49.308180+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"With coupling disabled, run many single-spin trials while deliberately misaligning the input polarization or detuning PC2 from its π setting; if the probability of +1 versus −1 departs from 50/50, or if the final-energy distribution of 64-spin chains shifts measurably, the topological symmetry protection on which the bias-free claim rests has failed.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the nonlinear topological symmetry protection mechanism (localized birefringent defect) on which the claim of bias-free spins rests."},{"cited_title":"G., Coen, S","cited_arxiv_id":null,"evidence_quote":"Establishes that spontaneous symmetry breaking in a Kerr resonator yields unbiased random binary states read out via stroboscopic intensity measurements, the basis for the spin encoding."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Original coherent Ising machine proposal based on degenerate optical parametric oscillators; defines the Ising Hamiltonian and phase encoding that the paper replaces with polarization."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates time-multiplexed optical parametric oscillator networks as coherent Ising machines, the architecture the polarization scheme extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Large-scale coherent Ising machine showing the phase-stabilization and post-selection challenges the paper aims to avoid."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fully programmable 100-spin coherent Ising machine with all-to-all connections; provides the FPGA feedback methods the paper cites for future all-to-all coupling."},{"cited_title":"& Wabnitz, S","cited_arxiv_id":null,"evidence_quote":"Establishes polarization multistability and instability in nonlinear dispersive ring cavities, the physical effect behind the polarization spins."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the symmetry-breaking analysis for driven Kerr resonators used to interpret the polarization spontaneous symmetry breaking bifurcation."},{"cited_title":"& Yamamoto, Y","cited_arxiv_id":null,"evidence_quote":"Supplies the performance comparison against classical neural networks and the scaling benchmark that the paper's time-to-solution data are compared with."}],"review_version":1}