{"id":"ecd40026-5a78-4d7f-ad10-2a4688fa37b5","arxiv_id":"2608.05293","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The maximum Schmidt rank representable by a degree-p polynomial decoder with K latent variables is exactly binom(K+p,p), so for p proportional to K the entangling power grows exponentially.","lead":"This paper introduces a new metric, called entangling power, that measures how much quantum entanglement a neural network can produce between two halves of a system, and derives an exact formula for polynomial decoders. The formula shows that a network with a modest latent space can represent states with exponentially large entanglement, something a linear decoder cannot do.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Exact Bell-state width K_min≈0.2938n is proven only for unconstrained encoders; the neural implementation uses K=n, so the exact constant is not yet tied to low-complexity neural networks.","rationale":"The reader's weakest_assumption identifies exactly the same issue: the converse proof assumes arbitrary encoders, which may not be realizable by low-complexity neural networks. I agree with that assessment. However, this concern does not invalidate the mathematical theorem as stated, because the theorem explicitly declares the encoders to be arbitrary. Moreover, the paper provides an explicit neural-network implementation with K=n and O(n) units that already demonstrates exponential entangling power, so the core claim of the abstract is supported independently of the exact constant. The remaining gap is that the exact minimal width K_min≈0.2938n is not tied to neural-computable encoders, and no construction or complexity argument is given for it. This is a genuine limitation but not a fatal flaw; the verdict should remain ACCEPT/UNCHANGED, ideally with an explicit caveat in the paper clarifying that the minimal-width result is for arbitrary encoders and may not be achievable by efficient neural networks. The proposed test would determine whether a simple family of encoders can realize the minimal width, providing concrete evidence about whether the concern lands.","tokens_in":10,"tokens_out":24787,"duration_ms":294232,"concrete_test":"For n=4,...,20, set D=2^n, p=n, and K=ceil(K_p(D)). Search for integer weights w_1,...,w_K modulo D such that the exponent set E={sum_j i_j w_j mod D : i_j>=0, sum_j i_j<=p} contains at least D distinct residues. If such weights exist, the Bell state can be represented with K=K_p(D) using encoders phi_A(x)=(omega^{w_1 N(x)},...,omega^{w_K N(x)}) of O(nK) circuit size, making the minimal-width construction efficiently computable. If no such weights are found for any tested n, that would support the concern that the exact K_min requires super-efficient encoders, and the paper should be revised to state that the constant 0.2938n applies only to the unrestricted-encoder model.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The converse direction of the theorem (Supplemental Theorem 1) proves that every f can be represented with K=K_p(D) latent variables by choosing encoder points a_x in C^K whose monomial evaluation vectors are linearly independent. These encoders are arbitrary maps; the proof gives no construction and no complexity bound. For the n Bell-pair state, the claimed exact minimum K_min=K_p(2^n)≈0.2938n for p=n therefore rests entirely on the unrestricted-encoder model. The paper's explicit neural-network implementation in the 'Neural-network implementation' section uses K=n (Eq. (10)), not K_min, and relies on a specially factored decoder. Thus the headline claim 'neural network with O(n) latent variables ... minimum latent width ... 0.2938n' conflates the general polynomial-decoder model (arbitrary encoders) with the neural-network model (efficiently computable encoders). The lower bound R≤N_p(K) and the explicit K=n construction are unaffected, but the exact constant 0.2938n for neural networks with polynomially computable encoders is an open question. This is a real soft spot because the paper's title and abstract emphasize neural networks, and the exact minimal width is a central quantitative result.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces an 'entangling power' E_p(K) for encoder-decoder representations of bipartite functions, where the decoder is a polynomial of total degree at most p in each latent vector and the latent space has dimension K. The central theorem states that, when the two input sets are large enough, the maximum Schmidt rank representable with K latent variables and degree-p polynomial decoders is exactly N_p(K) = binom(K+p, p). For the n-Bell-pair state, the authors derive the exact minimal latent width K_min = K_p(2^n), giving K_min ≈ 0.2938n for p=n, and K_min = 2^n−1 for p=1. The proof is based on a monomial expansion for the lower bound and on constructing indicator polynomials via a full-row-rank evaluation matrix for the upper bound. The paper also gives an explicit neural-network implementation of the Bell state with K=n using quadratic activations, and claims exponential entangling power of neural networks.","tokens_in":8692,"tokens_out":11797,"duration_ms":93591,"significance":"If the results stand, the paper gives a clean combinatorial characterization of the representational power of polynomial decoders, showing that nonlinear decoders can generate exponentially large Schmidt rank from a surprisingly small latent space. The main theorem is proven in a self-contained way: the lower bound follows directly from the monomial expansion, and the upper bound is an explicit interpolation construction with no fitted parameters. The asymptotic constants, including α_0 = 0.293815..., are derived in the Supplemental Material. The explicit Bell-state construction with K=n and O(n) quadratic-activation units is a concrete, checkable contribution. The notion of entangling power as e^{S_max} generalizes Schmidt rank in a natural way and may be useful for understanding neural quantum states, although the neural-network versus abstract-polynomial-decoder distinction needs care.","major_comments":[{"comment":"The exact minimal width K_min = K_p(2^n) ≈ 0.2938n for the Bell state is proven only for arbitrary encoders: the upper-bound construction chooses encoder points a_x in C^K whose monomial evaluation vectors are linearly independent, with no complexity bound or neural-network realizability. The only explicit neural-network implementation, Eq. (10), uses K=n, not K_min, and relies on a specially factored decoder. Therefore the exact constant 0.2938n is not yet established for neural networks with polynomially computable encoders; only K=n is. Since the title and abstract emphasize neural networks and the exact minimal width is a central quantitative claim, the authors should either provide an efficient encoder construction achieving K_min for the Bell state, or explicitly and prominently state that the exact constant applies to the abstract class of polynomial decoders with unrestricted encoders, while the neural-network result gives K=n.","section":"Polynomial decoders / Neural-network implementation (Eq. (25), Eq. (10), Supplemental Theorem 1)"},{"comment":"The theorem as stated says E_p(K)=N_p(K) 'provided D_A,D_B ≥ N_p(K)' and then lists K_p(R) ≤ K_min ≤ K_p(D) as a consequence, but the upper-bound proof actually requires the reverse condition N_p(K) ≥ D (the text says 'when N ≥ D'). The proof is correct for the two separate statements: (i) the maximal Schmidt rank over representable functions is N_p(K) when the input sets have size at least N_p(K); and (ii) every function on sets of size D is representable when N_p(K) ≥ D. The theorem statement should separate these two cases, because as written the stated condition does not logically yield the claimed upper bound K_min ≤ K_p(D).","section":"Polynomial decoders, Theorem statement and proof (Eqs. (12)-(20))"}],"minor_comments":[{"comment":"The asymptotic threshold equation (1+α)log(1+α)−α log α = 1 is correct only if log denotes base-2 logarithm; with natural logarithms the right-hand side should be ln 2. Please specify the logarithm base explicitly, since the main text also uses log in entropy expressions.","section":"Supplemental Material, Eqs. (38)-(39)"},{"comment":"The notation E = e^{S_max} is nonstandard and may confuse readers who expect 'entangling power' to be a rate or an entropy. Please define the normalization convention clearly and explain why the exponential of the maximum entropy is the natural quantity.","section":"Introduction, definition of entangling power (paragraph after Fig. 1 caption)"},{"comment":"The sentence 'this generally requires exponentially large complexity in g, as shown by Eq. (7)' is not a formal complexity statement; Eq. (7) merely exhibits one decoder with D summands. It would be helpful to state that the decoder's arithmetic complexity is O(D) in that construction, without claiming a lower bound.","section":"Maximally entangled example, Eq. (7)"}],"recommendation":"major_revision","confidential_remarks":"The mathematical core is sound and the paper is likely publishable after revision. The main issue is that the abstract and title present the exact minimal-width result (0.2938n) as a property of neural networks, whereas the proof only supports it for arbitrary encoders; the explicit neural implementation uses K=n. This is not a fatal flaw, but it is a load-bearing mismatch that must be corrected before acceptance. Please ask the authors to restate the theorem's two regimes and to qualify the neural-network claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper gives an exact formula for the entangling power of polynomial decoders: with K latent variables and degree p, the maximum Schmidt rank you can generate is binom(K+p,p), provided the two input sets are large enough. The proof is clean and self-contained: upper bound from counting monomials, lower bound by constructing indicator polynomials via a full-row-rank evaluation matrix. The Bell-state corollary is nice: p=n gives K_min ~ 0.2938 n, linear in n versus exponential Schmidt rank. That is a genuine result and the mathematics holds up. No fitting, no dependence on prior results, elementary but rigorous.\n\nThe soft spot, which the stress-test note flags correctly, is that the exact constant 0.2938 n is proven for arbitrary encoders, not for encoders you can actually compute with a small network. The converse theorem chooses latent points whose monomial evaluations are linearly independent; it gives no complexity bound on the encoder. The explicit neural-network construction in the paper uses K=n, not 0.2938n. So the abstract's phrasing 'modest resources' and the emphasis on neural networks slightly overstate what is proven. The lower bound R <= binom(K+p,p) is robust and applies to any encoder, so the exponential-in-p benefit of nonlinear decoders is real. But the exact minimal width for low-complexity neural encoders remains open.\n\nThat said, this is a legitimate theoretical contribution. The framework generalizing Schmidt rank to nonlinear decoders is useful, and the sandwich bound K_p(R) <= K_min <= K_p(D) is a nice tool. The paper deserves a serious referee. A good referee should ask the authors to separate the 'polynomial decoder with arbitrary encoder' results from the 'neural network' claims, and to state explicitly that the 0.2938n constant is existential. I would accept it with that revision.\n\nRecommendation: send to peer review. Bring to reading group if you want a clean example of expressibility bounds via counting.","headline":"A clean, exact entangling-power formula for polynomial decoders, with the caveat that the sharp Bell-state width constant is proven only for unconstrained encoders.","tokens_in":9195,"tokens_out":2015,"would_cite":true,"duration_ms":18246,"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":"This paper proves that the entangling power of a degree-p polynomial decoder with latent width K is exactly the binomial coefficient binom(K+p,p), making exponential entanglement accessible with a linear number of latent variables.","keywords":["entangling power","encoder-decoder neural networks","polynomial decoders","Schmidt rank","latent width","Bell states","entanglement entropy","quantum state representation"],"falsifier":"The claim is disproved by finding a single function f on finite sets with Schmidt rank strictly greater than binom(K+p,p) that still has an exact representation f = g(phi_A(x), phi_B(y)) with a degree-p polynomial decoder and latent width K, using unrestricted encoders; for instance, with K=1 and p=2 the bound is 3, so a 4x4 matrix of rank 4 representable this way would falsify the theorem.","tokens_in":8265,"feed_emoji":"⚛️","tokens_out":8267,"duration_ms":61770,"temperature":0.7,"pith_summary":"The paper introduces the entangling power of an encoder-decoder neural network: the maximum entanglement entropy (equivalently, Schmidt rank) that a network of given latent width and decoder complexity can generate between two halves of a system. For polynomial decoders of degree p, it proves the entangling power is exactly the number of monomials in K variables of total degree at most p, namely binom(K+p,p). This yields a sharp tradeoff: a linear decoder needs 2^n-1 latent variables for n Bell pairs, a degree-n decoder needs only about 0.2938n, and a degree at least 2^n-1 decoder needs just one. The result establishes that modest nonlinearity gives neural networks exponentially large entangling power, generalizing the Schmidt-rank notion of entanglement complexity.","feed_headline":"Polynomial decoders entangle exponentially with tiny latent width","feed_subtitle":"For n Bell pairs, a degree-n decoder needs just 0.29n latent variables instead of 2^n Schmidt coefficients.","key_machinery":"The central mechanism is the monomial evaluation map m(u)=(m_1(u),...,m_N(u)) from the latent space C^K to C^N, where the m_a are the monomials in K variables of total degree at most p. Its image spans C^N, so one can select encoder outputs whose evaluation vectors are linearly independent and solve the linear system that prescribes indicator polynomials q_x(u) with q_x($\\varphi$(x'))=delta_{xx'}. The decoder is then built as a sum over x,y of f(x,y) q_x(u) q_y(v), which both achieves the bound and, via expansion into a sum of N product functions, establishes the Schmidt-rank upper bound.","core_discovery":"On the paper's own terms, the discovery is a theorem with matching upper and lower bounds. For finite input sets X_A and X_B, any function f(x,y) represented as g(phi_A(x), phi_B(y)) with a degree-p polynomial decoder and K latent variables has Schmidt rank at most N_p(K)=binom(K+p,p); conversely, if both input sets have at least N_p(K) elements, every function admits such a representation. Hence the entangling power E_p(K), the maximum Schmidt rank obtainable, equals N_p(K). For n Bell pairs, where rank and dimension are both 2^n, the minimal latent width is exactly K_p(2^n), giving K_min = 2^n-1 for p=1, K_min ~ 0.2938n for p=n, and K_min = 1 for p >= 2^n-1. The paper also gives an explicit O(n)-unit, O(log n)-depth network implementing the degree-n Bell-state construction with quadratic activations.","pith_inferences":["The paper leaves open what happens if the encoders are restricted to affine maps or to networks of depth comparable to the decoder; the exact binomial bound may then fail, and the achievable entangling power would likely be governed by a different algebraic invariant.","A natural extension the authors do not pursue is the same counting argument for other decoder function classes, such as sparse polynomials or rational functions, where the relevant count of independent terms is no longer binomial.","The theorem reframes latent width as a measure of 'nonlinear rank' for classical bipartite functions, which could be a more appropriate complexity measure for neural network representations than matrix rank.","Because the Bell-state construction uses only a quadratic activation, it can be tested by direct simulation: build the O(n)-unit, O(log n)-depth network for small n and check that it reproduces the delta function exactly."],"forward_implications":["For n Bell pairs, the minimum latent width is exactly K_p(2^n), so a degree-n decoder needs about 0.2938n latent variables, collapsing an exponential Schmidt rank into a linear resource.","The entangling-power formula gives a rigorous tradeoff: increasing decoder degree p lowers the latent width needed, with p = 2^n-1 reducing it to a single latent variable.","For p=1, E_1(K)=K+1, recovering the near-linear cost of bilinear (Schmidt) representations, with the +1 arising from the constant term.","The explicit construction via indicator polynomials provides a decoder in the class P_p for any target function once encoder points are chosen, so the bound is achievable in full generality.","Any maximally entangled state whose unitary matrix has a polynomial-size arithmetic-circuit representation admits a compact polynomial decoder, not just the identity Bell state."],"supporting_citations":[{"why":"Supplies the Schmidt decomposition formalism that the paper generalizes to nonlinear decoders.","marker":"[2]"},{"why":"Provides the matrix-product-state and tensor-network context in which Schmidt rank appears as the resource for representing quantum states.","marker":"[3]"},{"why":"Gives the area-law scaling of entanglement that motivates why Schmidt rank is the standard measure of wavefunction complexity.","marker":"[4]"}],"fun_headline_variants":["Polynomial decoders pack exponential entanglement into small latent spaces","Exact entangling power: polynomial decoders deliver exponential rank","Tiny latent width, exponential entanglement: neural net decoders","Achieve exponential Schmidt rank with polynomial neural decoders"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The theorem assumes the encoders can be completely arbitrary maps from input configurations to latent vectors; the construction needs points in the latent space whose monomial evaluation vectors are linearly independent, and nothing guarantees a low-complexity neural network can produce those particular points.","fun_headline_variants_meta":{"raw":{"variants":["Polynomial decoders pack exponential entanglement into small latent spaces","Exact entangling power: polynomial decoders deliver exponential rank","Tiny latent width, exponential entanglement: neural net decoders","Achieve exponential Schmidt rank with polynomial neural decoders"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000304,"raw_usage":{"total_tokens":1707,"prompt_tokens":865,"completion_tokens":842,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":481,"completion_tokens_details":{"reasoning_tokens":774}},"tokens_in":481,"tokens_out":842,"duration_ms":8444,"temperature":1.0,"reasoning_tokens":774,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T15:54:12.658488+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The claim is disproved by finding a single function f on finite sets with Schmidt rank strictly greater than binom(K+p,p) that still has an exact representation f = g(phi_A(x), phi_B(y)) with a degree-p polynomial decoder and latent width K, using unrestricted encoders; for instance, with K=1 and p=2 the bound is 3, so a 4x4 matrix of rank 4 representable this way would falsify the theorem.","supporting_citations":[{"cited_title":"Eckart and G","cited_arxiv_id":null,"evidence_quote":"Supplies the Schmidt decomposition formalism that the paper generalizes to nonlinear decoders."}],"review_version":1}