{"id":"167a7af1-fdbc-417f-b834-896e92a381a7","arxiv_id":"2411.10295","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A self-interacting single-particle CBO process is shown to have its occupation measure converge polynomially to a unique invariant measure; the global-minimizer approximation remains unproven.","lead":"The paper introduces a single-particle, self-interacting variant of consensus-based optimization and proves that its time-averaged history converges to the unique invariant measure of a rescaled mean-field CBO process. The advertised guarantee that this limit approximates a global minimum is not established and rests on an unverified hypothesis.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Global-minimization claim fails: the bridge from Laplace limit (16) to the Dirac point-mass conclusion in §A.3 is mathematically invalid, and the α-uniform mass condition it would need is left unproven.","rationale":"The reader correctly identified the unverified α-uniform mass lower bound μ*_α(A_ε)≥C_ε as the weakest assumption connecting convergence to an invariant measure with global minimization. My stress-test goes one step further: the derivation that is supposed to use this assumption is itself invalid. The identity in §A.3 equating the normalization quotient with the integral of a point indicator is false for atomless invariant measures, and the claim that the Laplace limit forces the quotient to tend to 1 is false even when the Laplace limit holds. This matters because the abstract promises a global-minimizing invariant measure, while Theorem 3.5 only establishes convergence to an unidentified μ*_α. I do not object to the convergence result itself; its proof follows the framework of [4] and the cited propositions appear applicable under the stated assumptions on λ, σ, κ and the weight classes. The verdict should remain CONDITIONAL: the paper's core convergence-to-invariant-measure theorem is plausible, but the advertised global-minimization statement needs either a correct proof (including the missing mass condition) or a substantial softening of the abstract and introduction.","tokens_in":14214,"tokens_out":14989,"duration_ms":155372,"concrete_test":"Evaluate, for μ*_α=N(0,1), f(x)=x² and x*=0, the two quantities equated in §A.3: e^{-αf(x*)}/∫ω^α dμ*_α = √(1+2α) and ⟨η*_α,I_{x*}⟩ = 0. The first diverges as α→∞ while (16) holds, and the second is identically zero for every atomless μ*_α; this directly invalidates the displayed bridge and shows that the global-minimization step needs a genuinely new argument rather than a corrected constant.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 3.5 (with Corollary 3.8) plausibly establishes polynomial convergence of the occupation measures E^ϑ_t[Y] and E^ϑ_t[X] to the unique invariant measure μ*_α of the rescaled mean-field CBO (7). But the paper's advertised application—that this invariant measure approximates the global minimizer—is not a consequence of Theorem 3.5. The only bridge is the paragraph after (16) and the supplementary computation in §A.3. That bridge is broken in two places. First, the Laplace limit −(1/α)log Z_α → f(x*) only determines the exponential rate of Z_α; it does not imply e^{-αf(x*)}/Z_α → 1. Example: μ*_α=N(0,1), f(x)=x², x*=0 gives Z_α=1/√(1+2α); the Laplace limit holds, yet e^{-αf(x*)}/Z_α=√(1+2α)→∞. Second, the paper writes e^{-αf(x*)}/Z_α = ⟨η*_α,I_{x*}⟩. For an atomless μ*_α—the expected situation given the uniformly elliptic term (1/α)Id—the right-hand side is identically 0, while the left-hand side is generally nonzero; the equality silently drops μ*_α({x*}). The correct Laplace consequence is concentration in neighborhoods of x*, and that still requires the α-uniform lower bound μ*_α(A_ε)≥C_ε introduced after (16), which the authors explicitly say is 'undergoing work.' Since the abstract's 'approximates the global minimum' and the connection to Personal Best/CBO both rest on this unproven and currently misstated bridge, the central advertised claim is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a single-particle self-interacting CBO dynamics (6) with an occupation-measure consensus point, together with generalized weighted versions (9)-(10). It claims that both the self-interacting process and the rescaled mean-field CBO (7) converge, in expected occupation-measure Wasserstein distance, to a common invariant measure μ*_α at a polynomial rate (Theorem 3.5), that μ*_α is unique (Corollary 3.8), and that μ*_α approximates a global minimizer through a Laplace-principle argument, thereby connecting the model to CBO with Personal Best. The convergence-to-invariant-measure part is supported by the estimates in Lemma 3.2 and external results [4,20]; the global-minimization bridge is not established.","tokens_in":14625,"tokens_out":9681,"duration_ms":78971,"significance":"If the convergence theorem is correct, the paper provides a useful single-particle alternative to N-particle CBO and extends the self-interaction framework of [4] to a consensus-based dynamics. The polynomial rate (17), the explicit weight classes, and the uniqueness result are concrete contributions. The paper is also honest in flagging the unverified α-uniform mass condition. However, the advertised application to global optimization is not established: the arguments after (16) and in Appendix A.3 contain a mathematical error, and the needed mass condition is left as future work. The present manuscript therefore establishes convergence to an α-dependent invariant measure, not a global-minimizer approximation; the significance is conditional on either a proof of the missing mass condition or a careful reframing of the claims.","major_comments":[{"comment":"The derivation that η*_α approximates δ_{x*} is mathematically invalid. From the Laplace limit lim_{α→∞} −(1/α) log Z_α = f(x*) one can conclude that e^{−f(x*)}/Z_α^{1/α} → 1, but raising this ratio to the power α is not justified, so the conclusion lim_{α→∞} e^{−αf(x*)}/Z_α = 1 does not follow. In addition, the equality e^{−αf(x*)}/Z_α = ⟨η*_α, I_{x*}⟩ silently drops the point mass μ*_α({x*}); for an atomless μ*_α the right-hand side is zero while the left-hand side is generally non-zero. The paper itself states that the required α-uniform lower bound μ*_α(A_ε) ≥ C_ε is 'undergoing work' (paragraph after (16)). Since the abstract's global-minimization claim rests entirely on this bridge, that advertised claim is not established.","section":"Paragraph after Eq. (16) and Appendix A.3"},{"comment":"The abstract states that the dynamics 'converges to a unique invariant measure that approximates the global minimum'. Theorem 3.5 as proven only gives polynomial convergence of expected occupation measures E^ϑ_t[Y] and E^ϑ_t[X] to μ*_α in Wasserstein distance; it does not give convergence of the law of the process at fixed times, nor any information about the support of μ*_α. The statement E[X_∞] = κ m_α(μ*_α) in (15) is formal because X_∞ is not defined. The paper should either prove the global-minimizer statement under an explicit assumption or revise the abstract and Section 1 to advertise only convergence to the unique invariant measure.","section":"Abstract and Theorem 3.5"},{"comment":"The verification of Assumption (H1) of [20, Theorem 2.2] contains a constant mismatch. The display before the definition of ̃C_1 has coefficient 2σ²(1+κ) in the |x|² term and 2σ²(1+κ)κ C_1² in the ν(|·|²) term, but ̃C_1 is then defined with σ²(1+κ²) and ̃C_3 with σ²(1+κ)κ. Because ̃C_1 as defined is larger than the coefficient actually obtained, the inequality 2⟨b(x,ν),ν⟩+‖σ(x,ν)‖² ≤ −̃C_1|x|² + ̃C_2 + ̃C_3 ν(|·|²) does not follow as written. This is likely repairable by taking ̃C_1 = (2λ−λκ)−2σ²(1+κ) and ̃C_3 = λκC_1²+2σ²(1+κ)κC_1², but as it stands Proposition 3.4 relies on an unverified estimate.","section":"Appendix A.2, Proposition 3.4"}],"minor_comments":[{"comment":"Assumption 2.1 does not guarantee that the global minimizer x* exists; the paper should add an attainment condition (e.g., f coercive and continuous, or explicitly assume argmin f is nonempty) because x* appears in (16) and in the definition of A_ε.","section":"Assumption 2.1"},{"comment":"In Theorem 3.11, the condition 'ϑ∈Π(ε2)' is almost certainly a typo for 'ϑ∈Π_2(ε2)'; please correct it to match the notation used in Theorem 3.5.","section":"Theorem 3.11"},{"comment":"In Corollary 3.8, the application of Theorem 3.5 to a second invariant measure ̃μ*_α should state explicitly why ̃μ*_α ∈ P_{2,R} for the relevant R; this follows from the Lyapunov-type estimate but should be spelled out.","section":"Corollary 3.8"},{"comment":"The bound ∫_0^1 t^ε ∧ s^{-ε} ϑ_t(ds) ≤ ∫_0^1 t^{ε_2} ∧ s^{-ε_2} ϑ_t(ds) requires ε ≤ ε_2; since ε = ε_1 ∧ γε_2 ≤ ε_2 this is fine, but the monotonicity should be stated explicitly for the reader.","section":"Proof of Theorem 3.5, after Eq. (20)"},{"comment":"The phrase 'as we shall later see, E[X_{t=∞}]≈κ x*' is informal because X_∞ is not a random variable; please rephrase using the invariant measure μ*_α, e.g., 'the mean of μ*_α is close to κ x*'.","section":"Introduction, paragraph before Eq. (4)"}],"recommendation":"major_revision","confidential_remarks":"The paper's core mathematical result (Theorem 3.5 and Corollary 3.8) is likely correct and fits the journal's scope, but the advertised global-minimization guarantee is not proven and the formal derivation in §A.3 is incorrect. I recommend major revision rather than rejection because the gap is localized: the authors can either add the missing α-uniform mass condition as a theorem or honestly reframe the contribution as convergence of a self-interacting CBO to its unique invariant measure. The constant mismatch in Appendix A.2 should also be corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The real content of this paper is the polynomial convergence of the occupation measure of a single self-interacting particle to the unique invariant measure of a κ-rescaled mean-field CBO. That is new, and it is a nontrivial extension of the Du–Jiang–Li framework to a non-dissipative setting. The proof of Theorem 3.5 is structurally sound: Lemma 3.2 gives the needed one-sided inequality with explicit constants, and the rest is a careful application of [4] and [20]. Corollary 3.8, uniqueness of the invariant measure, follows cleanly. The paper is transparent about the regime λ > 8σ² and κ ≪ 1, and it does not hide the fact that the main step uses external machinery. I found nothing circular or fitted in the main argument.\n\nThe soft spot is exactly where the abstract overstates the result. The claim that the invariant measure 'approximates the global minimum' is not proven. The paper needs an α-uniform lower bound on μ*_α(A_ε), which the authors explicitly say is 'undergoing work.' Worse, the supplementary computation in §A.3, meant to close this gap, is mathematically wrong: the equality e^{-αf(x*)}/Z_α = ⟨η*_α, I_{x*}⟩ silently assumes an atom at x*, which an atomless μ*_α will not have. The Laplace limit only gives concentration in α-dependent neighborhoods, not point-mass convergence. This is not a minor typo; it is the only bridge from the invariant measure to global optimization, and it breaks. The connection to Personal Best rests on the same broken bridge.\n\nThat said, the broken bridge does not infect the main theorem. Theorem 3.5 and Corollary 3.8 stand on their own. The paper would be a solid contribution if the authors either prove the missing condition or simply drop the global-minimizer claim from the abstract and the formal part, presenting the result as convergence to a unique invariant measure of a rescaled CBO.\n\nWho should read this? People working on consensus-based optimization and on self-interacting diffusions will find the convergence theorem and the uniqueness corollary useful and citable. The flaws are in the packaging, not in the core proof.\n\nRecommendation: send this to a serious referee. A good referee will ask for the abstract to be corrected and the A.3 argument fixed or removed, but the substantive result merits the attention.","headline":"A solid and genuinely new convergence theorem for a single-particle CBO is undermined by an over-sold abstract and a wrong supplementary 'global minimizer' step, but the core result deserves a referee.","tokens_in":15133,"tokens_out":2374,"would_cite":true,"duration_ms":25431,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H10","65C35","37N40","90C26"],"pacs":[],"model":"deepseek-v4-flash","headline":"One trajectory can replace the CBO particle swarm: a single self-interacting diffusion converges to the same invariant measure as the mean-field consensus-based optimization process, at a polynomial rate.","keywords":["self-interacting diffusion","consensus-based optimization","invariant measure","global optimization","McKean–Vlasov process","mean-field limit","personal best","occupation measure"],"falsifier":"For a chosen objective f, simulate the frozen Markovian process (18) long enough to approximate μ*_α, then evaluate $-\\frac{1}{\\alpha}\\log\\int e^{-\\alpha f(x)}\\,\\mu^*_\\alpha(dx)$ for increasing α. If this quantity does not approach f(x*) as α→∞, the mass hypothesis fails and the advertised global-minimization guarantee collapses.","tokens_in":13975,"feed_emoji":"🎯","tokens_out":5527,"duration_ms":54874,"temperature":0.7,"pith_summary":"This paper introduces a self-interacting version of consensus-based optimization (CBO): one particle whose drift and diffusion depend on its own past occupation measure, not on a swarm of N interacting particles. The authors try to establish that this single-particle process and the mean-field CBO process share the same unique invariant measure, and that both converge to it polynomially fast. If true, running one trajectory replaces simulating an interacting particle system, and the consensus point built from the trajectory concentrates near a global minimizer as the parameter α grows. The global-minimization part is conditional on an unverified mass assumption on the invariant measure near the minimizer, which the paper states is still being worked on.","feed_headline":"One trajectory can replace the CBO particle swarm","feed_subtitle":"A self-interacting single particle converges to the same invariant measure as the mean-field CBO, polynomially.","key_machinery":"The load-bearing object is the occupation measure $E_t[Y] = \\frac{1}{t}\\int_0^t \\delta_{Y_r}\\,dr$, used in place of the law $L_t[X]$ inside the consensus point $m_\\alpha(\\cdot) = \\frac{\\int x e^{-\\alpha f(x)}\\,\\cdot(dx)}{\\int e^{-\\alpha f(x)}\\,\\cdot(dx)}$. The model also uses the rescaling $0<\\kappa\\ll1$ in the drift and the nondegenerate diffusion $\\sigma(\\alpha^{-1} I_d + D(Y_t-\\kappa m_\\alpha(E_t[Y])))$, which together rule out the Dirac invariant measures that plague the standard CBO dynamics. The convergence proof compares $Y$ and the mean-field $X$ with an auxiliary Markovian SDE in which $m_\\alpha(\\mu^*_\\alpha)$ is frozen; a Gronwall-type estimate and weight classes $\\Pi_1(\\varepsilon),\\Pi_2(\\varepsilon)$ control the expected Wasserstein distance between weighted occupation measures.","core_discovery":"The central claim is that the rescaled mean-field CBO process (7) has a unique invariant measure μ*_α, and the single-particle self-interacting process (6) converges to it in the sense of expected squared Wasserstein-2 distance of its occupation measure, at a polynomial rate $t^{{-ε}}$. The proof couples both processes to an auxiliary Markovian process (18) with a frozen consensus point m_α(μ*_α), whose weighted occupation measure is shown to converge using comparison estimates for occupation measures. Because the same invariant measure is shared, the occupation measure of one trajectory can substitute for the N-particle empirical measure in the long-time limit. With an additional unverified Laplace-type mass condition, the consensus point built from the trajectory approximates a global minimizer as α→∞.","pith_inferences":["The unverified mass assumption is the only bridge from convergence to an invariant measure to convergence to a global minimizer; if it fails, the optimization guarantee weakens to convergence to some invariant measure, not necessarily concentrated near the minimum.","Since the method only uses moments through the consensus point, it may extend to nonsmooth objectives satisfying the growth and Lipschitz-type conditions; a natural test is to run the single-particle scheme on standard high-dimensional nonconvex benchmarks.","A practical implementation would need a finite-memory approximation of the occupation measure and a discretization scheme; the paper does not analyze the resulting discretization error, which is a natural next step.","The connection to CBO with Personal Best suggests that memory-based optimizers could be analyzed through the same occupation-measure comparison, potentially giving convergence rates for other trajectory-weighted consensus algorithms."],"forward_implications":["The single-particle dynamics (6) can be used in place of the N-particle system (5) for long-time approximation of the CBO invariant measure, eliminating the need for N→∞.","Both the self-interacting and mean-field CBO processes converge to the same unique invariant measure at polynomial rate, strengthening the theoretical basis of CBO with Personal Best.","The rescaled mean-field CBO has a unique invariant measure even though it does not satisfy the usual dissipativity assumption, extending the class of McKean–Vlasov SDEs with uniqueness.","For large α, the consensus point built from the trajectory is expected to approximate $\\kappa x_*$, so a single trajectory can approximately locate a global minimizer.","The multi-particle analogue inherits a rate combining particle number and time, giving a finite-time, finite-ensemble guarantee for the empirical occupation measure."],"supporting_citations":[{"why":"Defines the original consensus-based optimization model whose McKean–Vlasov dynamics this paper modifies.","marker":"[17]"},{"why":"Introduces the Personal Best consensus term, the same object as the self-interacting occupation-measure consensus, and supplies well-posedness arguments.","marker":"[18]"},{"why":"Provides the weight classes Π1, Π2 and the comparison estimates for occupation measures that drive the polynomial convergence proof.","marker":"[4]"},{"why":"Supplies the existence theorem for stationary distributions of distribution-dependent SDEs used to obtain the invariant measure μ*_α.","marker":"[20]"},{"why":"Provides the Laplace-principle lemma that connects the invariant measure to the global minimizer under the mass assumption.","marker":"[14]"},{"why":"Establishes the Lipschitz and growth estimates for the consensus map m_α used throughout the estimates.","marker":"[10]"},{"why":"Gives the analytical framework for well-posedness of the mean-field CBO dynamics, used to verify the SDE conditions.","marker":"[1]"}],"fun_headline_variants":["Self-interacting single particle matches CBO's invariant measure","One particle suffices: self-interacting CBO convergence","Self-interacting CBO: unique measure, polynomial rate","Single trajectory replaces swarm in long-time CBO limit","Self-interacting particle converges to CBO's unique measure"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The only unproved step connecting the invariant measure to the global minimum is that, as α grows, the invariant measure keeps a fixed amount of mass on the set where f is within ε of its best value; the paper says verifying this is ongoing work.","fun_headline_variants_meta":{"raw":{"variants":["Self-interacting single particle matches CBO's invariant measure","One particle suffices: self-interacting CBO convergence","Self-interacting CBO: unique measure, polynomial rate","Single trajectory replaces swarm in long-time CBO limit","Self-interacting particle converges to CBO's unique measure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000832,"raw_usage":{"total_tokens":3541,"prompt_tokens":760,"completion_tokens":2781,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":376,"completion_tokens_details":{"reasoning_tokens":2701}},"tokens_in":376,"tokens_out":2781,"duration_ms":19476,"temperature":1.0,"reasoning_tokens":2701,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T19:47:08.263377+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a chosen objective f, simulate the frozen Markovian process (18) long enough to approximate μ*_α, then evaluate $-\\frac{1}{\\alpha}\\log\\int e^{-\\alpha f(x)}\\,\\mu^*_\\alpha(dx)$ for increasing α. If this quantity does not approach f(x*) as α→∞, the mass hypothesis fails and the advertised global-minimization guarantee collapses.","supporting_citations":[{"cited_title":"Pinnau, C","cited_arxiv_id":null,"evidence_quote":"Defines the original consensus-based optimization model whose McKean–Vlasov dynamics this paper modifies."},{"cited_title":"Totzeck and M.-T","cited_arxiv_id":null,"evidence_quote":"Introduces the Personal Best consensus term, the same object as the self-interacting occupation-measure consensus, and supplies well-posedness arguments."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the weight classes Π1, Π2 and the comparison estimates for occupation measures that drive the polynomial convergence proof."},{"cited_title":"Zhang, Existence and non-uniqueness of stationary distributions for distribution dependent SDEs , Electronic Journal of Probability 28 (2023), 1–34","cited_arxiv_id":null,"evidence_quote":"Supplies the existence theorem for stationary distributions of distribution-dependent SDEs used to obtain the invariant measure μ*_α."},{"cited_title":"Huang, J","cited_arxiv_id":null,"evidence_quote":"Provides the Laplace-principle lemma that connects the invariant measure to the global minimizer under the mass assumption."},{"cited_title":"A Carrillo, Y.-P","cited_arxiv_id":null,"evidence_quote":"Gives the analytical framework for well-posedness of the mean-field CBO dynamics, used to verify the SDE conditions."}],"review_version":1}