{"id":"04d0dd03-d598-4f80-b549-56f716006ff8","arxiv_id":"2412.02396","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A single compact Lefschetz thimble restores correct complex Langevin convergence, and a Dyson-Schwinger bias correction recovers the original expectation values.","lead":"The authors show that complex Langevin simulations converge correctly when the regularized system has a single compact Lefschetz thimble, and they introduce a Dyson-Schwinger based bias correction that removes the effect of the regularization. This makes the method work for several toy models where direct complex Langevin failed, and it points toward a heuristic principle for designing stabilization kernels.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Periodic regularization (Eq. 35) makes the effective drift discontinuous; the Nagata criterion, proven for holomorphic actions, is applied without justification to this non-holomorphic setting, so the claimed correct convergence rests on direct exact-value checks, not on the stated criterion.","rationale":"The reader's weakest assumption identifies the same core issue: the Nagata et al. criterion is applied to non-holomorphic effective actions arising from periodic regularization without proving its validity. My analysis sharpens this: the periodic continuation makes the drift discontinuous, which is a stronger obstruction than mere non-holomorphicity, since the standard CL justification relies on smooth analytic drifts. This is load-bearing because the paper uses the criterion to claim correct convergence for the regularized models, and if the criterion is inapplicable, that claim needs to fall back entirely on the direct exact-value comparisons, which are strong but limited to specific parameter sets. The proposed test directly addresses this by checking the Fokker-Planck solution with the true discontinuous drift, which would conclusively show whether boundary terms invalidate the approach. I do not see a more fundamental flaw: the bias-correction method is algebraically sound and validated numerically, the thimble conjecture is explicitly stated as conjectural with supporting evidence, and the direct exact comparisons for all models provide genuine independent support. The paper is careful in its claims and the limitation of Appendix B honestly notes the difficulty of generalizing to lattice field theories. Therefore, the appropriate verdict remains CONDITIONAL, unchanged from the reader's assessment, with the expectation that the non-holomorphic correctness concern be addressed before the criterion-based conclusions are taken as established.","tokens_in":35034,"tokens_out":13824,"duration_ms":143683,"concrete_test":"Solve the Fokker-Planck equation for the regularized cosine model (beta = 0.5, r = 0.5) on the domain (-pi, pi) x R with the actual discontinuous drift from Eqs. (34)-(35), imposing zero-flux boundary conditions at Re z = ±pi. Compare the stationary expectation values <O>_CL with the exact integrals of O over [-pi, pi] with weight rho + R. If they match and the probability flux across the periodic boundaries vanishes, the discontinuity is benign; if not, the p(u) criterion is insufficient for this non-holomorphic setting and the bias correction may be invalid. Repeat for r = 0.1 and r = 2 to test parameter dependence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central evidence that CL converges correctly for the regularized models uses the Nagata et al. correctness criterion [57] (exponential decay of the drift-magnitude distribution p(u)) as an observable-agnostic diagnostic (Figs. 5 and 11). However, the periodic continuation in Eq. (35) makes the regularized weight and hence the drift discontinuous across the lines Re z = ±π for Im z ≠ 0. The Nagata criterion is derived for holomorphic actions, where the drift is analytic and boundary terms can be controlled through decay of p(u). No argument is given that the criterion remains sufficient when the drift has such discontinuities, and the paper's only defense is the observation that the simulated boundary segments are repulsive. This is an empirical claim, not a proof, and it is tested only for the specific parameters chosen. If the discontinuity generates hidden boundary contributions, then CL for the regularized weight may not reproduce the regularized integral, and the bias correction in Eqs. (26)-(29) would be invalid. The direct comparisons with exact regularized expectation values (Fig. 4, Tables 6 and 10) provide independent support and are currently reassuring, but the paper's repeated assertion that the correctness criterion is satisfied is not a reliable foundation in this non-holomorphic setting. A dedicated check of the Fokker-Planck solution with the discontinuous drift would settle whether this concern actually lands.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an additive weight-regularization scheme for complex Langevin (CL) simulations: an extra term R(z) is added to the original weight so that the regularized theory develops a single relevant compact Lefschetz thimble, which the authors argue restores correct CL convergence. A bias-correction formula, Eq. (26), relates expectation values of the regularized theory to those of the original theory through an observable-independent ratio Q, which is extracted using Dyson-Schwinger equations of the original theory. The method is applied to the complex cosine model and to SU(2) and SU(3) Polyakov chain models, with exact integration as a benchmark. The authors find that direct CL fails in the unregularized cases, that regularized CL reproduces the exact regularized expectation values, and that the bias-corrected values agree with the original exact values, typically at the sub-percent level. They also examine the Nagata et al. drift-magnitude criterion p(u) and report exponential decay for the regularized one-dimensional models.","tokens_in":35361,"tokens_out":9611,"duration_ms":109477,"significance":"If the results hold, the paper offers a constructive remedy for a known failure mode of CL and provides some of the clearest numerical evidence to date connecting thimble structure to CL convergence. The main technical novelty is the Dyson-Schwinger-based bias correction, which addresses a limitation of the earlier weight-regularization proposals in Refs. [68-70]. Strengths of the manuscript include exact benchmarks for every model, Fokker-Planck checks for the one-dimensional models, explicit p(u) diagnostics, and a candid discussion of limitations, including the argument in Appendix B that additive regularization is impractical for lattice field theories. The central conjecture, that a single relevant compact thimble is sufficient for correct convergence, is appropriately presented as numerical evidence rather than as a theorem. The main weaknesses are the uncritical use of a holomorphic correctness criterion for the non-holomorphic periodic regularization and an erroneous table in the SU(2) link-model results.","major_comments":[{"comment":"The periodic continuation in Eq. (35) makes the regularized weight and hence the drift discontinuous across the lines Re z = ±π for Im z ≠ 0, not merely non-differentiable. The correctness criterion of Ref. [57] is stated in Sec. 2.2 for holomorphic actions, where boundary terms can be controlled through the decay of p(u). The paper applies this criterion to the non-holomorphic regularized cosine model without an argument that the criterion remains sufficient when the drift has such discontinuities. The observation that the boundary segments are repulsive is empirical and parameter-dependent. This matters because Section 7 and the abstract use the p(u) criterion as a principal justification that the regularized CL is correct. I recommend either supplying a boundary-flux or Fokker-Planck check for the discontinuous drift, or explicitly demoting the p(u) check to supportive evidence and relying on the exact-value comparisons of Fig. 4 and Table 2 as the primary correctness test.","section":"Sec. 4, Eq. (35); Sec. 4.2, Fig. 5"},{"comment":"Table 11, titled \"Bias-corrected results for the regularized SU(2) Polyakov chain link model,\" lists the exact values of the original theory as 0.275920, 0.268989, 0.142204, etc., which are identical to the regularized exact values in Table 10. The original-theory exact values from Table 9 are 0.339351, 0.212100, 0.147098, etc. As printed, the table contradicts the claim in Fig. 14 that the bias correction recovers the original theory, and it undermines one of the central numerical benchmarks. The exact column must be corrected, and the bias-corrected CL column should be checked to ensure it was not accidentally copied from Table 10.","section":"Appendix E, Table 11"}],"minor_comments":[{"comment":"The sentence \"It appears that the solution seems to be unaffected by these results\" should refer to \"these discontinuities\" rather than \"these results,\" and the text should state explicitly that R is discontinuous, not merely non-differentiable, at the periodicity boundaries for nonzero imaginary part.","section":"Sec. 4, after Eq. (35)"},{"comment":"The dashed line ∝ u^{-2} is presented as a power-law reference, but no fitting range is given for the unregularized tail, and the regularized case is described only visually as exponential; please specify the range and, if possible, a fitted decay rate.","section":"Sec. 4.2, Fig. 5"},{"comment":"For the SU(3) model the p(u) criterion is not shown; the conclusion of correct convergence there rests entirely on the exact-value comparisons in Tables 14-15. This is acceptable, but it should be stated explicitly so the reader does not infer that the criterion was checked in the non-Abelian case.","section":"Sec. 6"},{"comment":"The statement that CL converges correctly when a single relevant compact thimble contributes should be qualified as holding for the models and parameter sets studied, since the paper itself presents this as an empirical conjecture rather than a proven criterion.","section":"Sec. 7 and Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the scope of a lattice-field-theory journal and the numerical benchmarks are substantial. The two issues that require attention before publication are the erroneous exact column in Table 11 and the unjustified transfer of the holomorphic Nagata et al. correctness criterion to the non-holomorphic periodic regularization in Section 4. Both are fixable through table correction and a revised interpretation or additional numerical checks. I see no citation or disclosure problems; the relationship to Refs. [68-70] is described transparently."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should read this one. The new piece is the Dyson-Schwinger bias correction (their Method 2): instead of leaving the additive weight regularization biased, they extract the partition-function ratio Q from an observable whose expectation value vanishes on the original theory, then correct any observable. That works across the cosine model, the reduced SU(2) chain, the SU(2) link model, and the SU(3) chain, with sub-percent agreement against exact values in every case. The SU(3) application at imaginary coupling is a real step beyond the earlier regularization papers [68–70]. They also give every model an exact benchmark, check the p(u) decay criterion, and solve the Fokker-Planck equation independently for the 1D models. Appendix A's counterexample, where multiple compact thimbles still violate the criterion, is honest and strengthens the central conjecture.\n\nThe soft spots are real but not fatal. The periodic regularization in Eq. (35) makes the drift discontinuous across Re z = ±π, and the Nagata criterion was derived for holomorphic actions. The stress-test concern lands: the p(u) plots alone do not establish correct convergence in this non-holomorphic setting. The paper's own defense—that the boundary segments are repulsive—is empirical and not a proof. But the direct comparisons of regularized CL against exact regularized values (Fig. 4, Tables 6 and 10) are independent support, and they are plentiful. So the correctness claim for these specific parameters rests on the exact-value checks, not on p(u); that is weaker than the text sometimes implies, but the evidence is there.\n\nAppendix B concedes that additive regularization becomes intractable for realistic lattice field theories, so the practical scope is toy models. That is a real limitation, though the paper states it plainly. The lack of a code/data release is a minor annoyance; the numerical method is simple enough that reproduction is feasible, but a repository would help. The single-compact-thimble claim is explicitly a conjecture, and they do not overstate it.\n\nVerdict: this deserves a serious referee. It is a solid, mostly honest numerical study that will be useful to anyone working on complex Langevin or thimble methods. A referee should ask for a more careful discussion of the non-holomorphic correctness criterion and, ideally, code release, but the core results look right.","headline":"A genuinely useful bias-correction idea and the cleanest numerical evidence yet for the single-compact-thimble criterion, with a real but non-fatal formal gap around the non-holomorphic periodic regularization.","tokens_in":35882,"tokens_out":1721,"would_cite":true,"duration_ms":22965,"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":"The paper claims that complex Langevin simulations converge correctly when the system has a single relevant compact Lefschetz thimble, and that a weight regularization plus Dyson-Schwinger bias correction can create that structure and…","keywords":["complex Langevin","sign problem","Lefschetz thimbles","weight regularization","Dyson-Schwinger equations","bias correction","Polyakov chain","stochastic quantization"],"falsifier":"Take the regularized cosine model of Eq. (34), change the periodic continuation so that the singularities of the effective action lie inside the integration box rather than on its boundary, and measure both the drift-magnitude distribution $p(u)$ and the expectation values against exact integration: a power-law tail in $p(u)$ together with correct expectation values, or exponential decay with wrong values, would show that the single-compact-thimble condition is neither necessary nor sufficient.","tokens_in":34860,"feed_emoji":"🎲","tokens_out":8964,"duration_ms":93960,"temperature":0.7,"pith_summary":"The paper tries to establish a working rule for when Complex Langevin (CL), a stochastic cure for the numerical sign problem, can be trusted: in every model they test, CL converges correctly when the theory's Lefschetz thimble structure reduces to one relevant compact thimble, and fails when it does not. The authors turn this rule into a method by adding a weight regularization that reshapes the thimbles into that single compact form, then remove the regularization bias with Dyson-Schwinger equations. This restores correct expectation values in the complex cosine model and in SU(2) and SU(3) Polyakov chains at couplings where direct CL fails. The paper also warns that additive regularizations are unlikely to scale to full lattice field theories, so the lasting value is the thimble-based design principle rather than the specific regularization.","feed_headline":"Complex Langevin converges when one compact thimble dominates","feed_subtitle":"A weight term reshapes thimbles; Dyson-Schwinger equations remove the bias.","key_machinery":"The load-bearing object is the Lefschetz thimble: for a critical point $z_\\sigma$ of the action, the thimble is the contour of steepest ascent along which the imaginary part of the action stays constant, and it is called compact when it does not escape to imaginary infinity but terminates at zeros or singularities of the weight. The regularization is the additive modification $\\tilde{\\rho}(z)=\\rho(z)+r\\,G(z)+R_0$, with $G$ positive on the real integration domain and chosen so that the regularized weight has exactly one relevant compact thimble for sufficiently large $|r|$. The bias it introduces is removed through the affine identity $\\langle O\\rangle_\\rho = \\langle O\\rangle_{\\tilde{\\rho}} + Q\\,(\\langle O\\rangle_{\\tilde{\\rho}} - \\langle O\\rangle_R)$, where the ratio $Q=Z_R/Z_\\rho$ is extracted without computing the problematic partition function by using Dyson-Schwinger equations, which provide observables whose expectation values vanish in the original theory. Correctness of the regularized simulations is checked with the criterion that the distribution $p(u)$ of the drift magnitude decays faster than any power law.","core_discovery":"The central claim, stated in Section 7, is that complex Langevin tends to converge correctly when the system has a single relevant, attractive critical point and the relevant thimble extending from it is compact; the authors offer this as a conjecture supported by all their simulations. They test it by manually regularizing the weight of systems that violate the condition, producing modified models whose thimble structure is single and compact, and then correcting for the modification. The correction works because the regularized expectation value is an affine function of the original one, and the only unknown, the ratio of the two partition functions, is fixed by the Dyson-Schwinger equations of the original theory. The numerical results show that the bias-corrected expectation values agree with exact integration for the complex cosine model and for SU(2) and SU(3) Polyakov chains in parameter regions where unregularized complex Langevin diverges or gives wrong values.","pith_inferences":["If the single-compact-thimble condition holds beyond the toy models, it provides an a priori diagnostic: one could scan the thimble structure of a proposed kernel or regularization before running expensive CL simulations.","The DSE-based bias correction is not limited to the additive scheme; any modification whose expectation values relate affinely to the original theory, including multiplicative reweightings when overlap permits, could use the same $Q$-extraction strategy.","A testable next step is to apply the same construction to a genuinely higher-dimensional model, such as a complex $\\phi^4$ theory in one spatial dimension, where the thimble structure can still be computed numerically, to see whether the compact-single-thimble criterion remains sufficient outside zero-dimensional integrals.","Because the paper's correctness checks rely on the Nagata criterion, which is trusted in the non-holomorphic periodic setting, the strongest version of the claim is only as solid as that criterion in that setting."],"forward_implications":["CL can be made to converge correctly in models where it previously failed by constructing a regularization whose thimble structure is a single compact relevant thimble and then bias-correcting with Dyson-Schwinger equations.","The bias-correction formula removes the old obstacle of computing the partition function of the original theory; only regularized expectation values and the ratio $Q$ from Dyson-Schwinger equations are needed.","A single compact relevant thimble can be used as a design principle: positive-definite regularization terms that vanish on the boundary of the integration domain, often tied to zeros of the Haar measure, produce such structures in gauge models.","Direct additive regularization for full lattice field theories is impractical because it becomes non-local or the bias correction requires exponentially many subset terms, so the practical route lies in kernels or multiplicative regularizations.","Multiple compact relevant thimbles are still not enough: the paper's Appendix A shows that CL fails even when all contributing thimbles are compact, strengthening the single-thimble condition."],"supporting_citations":[{"why":"Supplies the criterion of correctness based on the decay of the drift-magnitude distribution, used throughout as the observable-independent check of convergence.","marker":"[57]"},{"why":"Provides the analytic stationary distribution of the unregularized cosine model and the previously stated conjecture linking CL correctness to a single relevant thimble.","marker":"[67]"},{"why":"Previously studied the SU(2) Polyakov chain and connected CL convergence or failure to the thimble structure, including non-compact thimbles.","marker":"[65]"},{"why":"Introduced the additive weight regularization idea that the paper revisits and extends with a new bias-correction scheme.","marker":"[68]"},{"why":"Proposed the multi-modification method that the paper's first approach to extracting the ratio $Q$ resembles.","marker":"[70]"},{"why":"Demonstrated wrong convergence of the unregularized cosine model via boundary terms, providing the baseline failure that the regularization must fix.","marker":"[58]"},{"why":"Gave semi-classical evidence that multiple dominant critical points cause CL to fail, supporting the claim that a single relevant thimble is required.","marker":"[66]"},{"why":"Documents CL failure for the Polyakov chain without stabilization and provides the gauge-cooling technique used in the simulations.","marker":"[73]"}],"fun_headline_variants":["Reshape thimbles to rescue Complex Langevin","Single compact thimble fixes Complex Langevin","Bias-corrected thimble regularization tames CL","Thimble-inspired weights rescue Complex Langevin","One thimble + bias fix makes CL converge"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on the criterion that complex Langevin converges correctly whenever the distribution of drift magnitudes $p(u)$ decays faster than any power law, and on that criterion remaining valid when the regularized effective action is not holomorphic because the regularization is periodically continued; if that criterion is not sufficient in these non-holomorphic settings, the conclusion of correct convergence is not established.","fun_headline_variants_meta":{"raw":{"variants":["Reshape thimbles to rescue Complex Langevin","Single compact thimble fixes Complex Langevin","Bias-corrected thimble regularization tames CL","Thimble-inspired weights rescue Complex Langevin","One thimble + bias fix makes CL converge"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000277,"raw_usage":{"total_tokens":1630,"prompt_tokens":904,"completion_tokens":726,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":520,"completion_tokens_details":{"reasoning_tokens":649}},"tokens_in":520,"tokens_out":726,"duration_ms":7872,"temperature":1.0,"reasoning_tokens":649,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T23:29:55.244747+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the regularized cosine model of Eq. (34), change the periodic continuation so that the singularities of the effective action lie inside the integration box rather than on its boundary, and measure both the drift-magnitude distribution $p(u)$ and the expectation values against exact integration: a power-law tail in $p(u)$ together with correct expectation values, or exponential decay with wrong values, would show that the single-compact-thimble condition is neither necessary nor sufficient.","supporting_citations":[{"cited_title":"An improvement in complex Langevin dynamics from a view point of Lefschetz thimbles","cited_arxiv_id":"1508.04231","evidence_quote":"Introduced the additive weight regularization idea that the paper revisits and extends with a new bias-correction scheme."},{"cited_title":"Modifying partition functions: a way to solve the sign problem","cited_arxiv_id":"1709.05806","evidence_quote":"Proposed the multi-modification method that the paper's first approach to extracting the ratio $Q$ resembles."},{"cited_title":"Scherzer, E","cited_arxiv_id":null,"evidence_quote":"Demonstrated wrong convergence of the unregularized cosine model via boundary terms, providing the baseline failure that the regularization must fix."},{"cited_title":"Controlling complex Langevin dynamics at finite density","cited_arxiv_id":"1303.6425","evidence_quote":"Documents CL failure for the Polyakov chain without stabilization and provides the gauge-cooling technique used in the simulations."}],"review_version":1}