{"id":"36f0198e-c996-4dd8-85b5-0332f6aafdd3","arxiv_id":"2603.00940","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For two or three copies of a partially entangled two-qubit state, distilling nonlocal correlations directly can yield a higher CHSH violation than first distilling entanglement.","lead":"This preprint compares two ways to improve noisy entangled states: distilling entanglement into a Bell state versus directly distilling nonlocal correlations. It reports that for two or three copies, the direct nonlocality route can give a higher CHSH violation, sharpening the practical distinction between entanglement and nonlocality.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"n=3 branch of Theorem 1 depends on an unproven attainability claim for Eq. (3); the paper's own Sec. II-B limits the corresponding measurements to numerical verification.","rationale":"The reader's weakest assumption identifies the same load-bearing concern. The central claim has two branches; the n=2 branch is supported by the known single-copy optimum and the explicit observables in Eq. (4). The n=3 branch is not, because the value used for nonlocality distillation is only numerically verified. This is explicitly acknowledged in the manuscript, so it is not a manufactured objection. The n=3 attainability gap is decisive for the theorem as stated, and the verdict should remain CONDITIONAL: the paper can be fixed either by proving attainability or by restricting the theorem to n=2 and stating the n=3 result as a numerical observation. No stronger concern that would overturn the n=2 result was identified.","tokens_in":8441,"tokens_out":37001,"duration_ms":302610,"concrete_test":"Check Liang and Doherty (2006) to determine whether Eq. (3) for n=3 is proven achievable by an explicit set of collective measurements or is only an upper bound. Concretely, attempt to construct local observables A_i, B_i on three copies of |Ψ> at, say, p=0.75 that realize the value 2[2√2 t + r√(1+4t)] used in the proof. If no such observables can be provided, the n=3 branch of Theorem 1 must be weakened to an upper-bound statement or removed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The n=3 claim in Theorem 1 is established by substituting V_ND from Eq. (3) into the proof inequality. But Eq. (3) is introduced as the 'best known CHSH bound' from Liang and Doherty, and the paper immediately states that for n=3 'the corresponding measurements have only been verified numerically' (Sec. II-B). A numerical check is not a proof that the value is attainable; if Eq. (3) is only an upper bound on the CHSH value achievable by collective measurements, then V_ED ≤ V_ND only shows that the optimal entanglement-distillation value is no larger than a possibly unreachable upper bound. It does not show that nonlocality distillation can attain the higher value. The n=2 case does not share this gap, since for n=2 Eq. (3) reduces to the single-copy Gisin-Peres value with the explicit observables in Eq. (4). Therefore the n=3 branch of the central claim is unsupported as written. This is a missing proof, not a disagreement with consensus; the same limitation is stated in the manuscript itself.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper compares nonlocality distillation with entanglement distillation for the task of maximizing CHSH violation from n copies of a bipartite state. For pure states of the form |Ψ⟩ = √p|ψ⟩ + √(1−p)|ϕ⟩, Theorem 1 claims that for n=2 and n=3 nonlocality distillation achieves a higher CHSH value than the optimal entanglement distillation protocol. Theorem 2 makes a similar claim for a mixed-state family ρ = p|ψ⟩⟨ψ| + (1−p)|ϕ⟩⟨ϕ| for n=2, and provides a (non-tight) bound for n=3. The paper also includes a resource-estimation comparison for the n=2 pure-state case, finding that nonlocality distillation requires fewer logical qubits, lower logical depth, and fewer T states than entanglement distillation.","tokens_in":8709,"tokens_out":13644,"duration_ms":125963,"significance":"The central question — whether nonlocality distillation can beat entanglement distillation when only a few copies are available — is interesting and conceptually relevant to the distinction between entanglement and nonlocality as resources. If the n=2 pure-state claim holds, it provides a concrete small-copy regime where directly measuring collective nonlocality outperforms first concentrating a Bell state, even though the latter uses communication. The paper also makes an explicit attempt at resource estimation, which is a useful addition for practical relevance. However, the manuscript as written contains several load-bearing algebraic and logical gaps that prevent the claims from being accepted as proven; the n=3 branch relies on an unproven attainability statement, and the comparison is made against a protocol that is optimal for Bell-state probability, not necessarily for CHSH.","major_comments":[{"comment":"The displayed formulas for V_ED are algebraically incorrect and produce unphysical values. For n=2, the formula V_ED = 2 + 4(2√2−2)t + 2((√p+√(1−p))/√2)^4 gives V_ED=4 at p=0.5, while the state is then product and the true CHSH value is 2. For n=3, the analogous formula gives values above 2√2 in the stated intervals (e.g., at p=0.9 it exceeds 3.7). These errors invalidate the proof of Theorem 1 as written. The authors must re-derive the expressions for p_succ and V_ED and re-check the claimed inequalities.","section":"Section III, proof of Theorem 1"},{"comment":"The n=3 branch of Theorem 1 uses V_ND from Eq. (3) as an attained value, but the paper states in Sec. II-B that for n=3 the corresponding measurements 'have only been verified numerically.' If Eq. (3) is only an upper bound, the inequality V_ED ≤ V_ND only shows that the optimal entanglement-distillation value does not exceed a possibly unreachable upper bound; it does not prove that nonlocality distillation attains the higher value. To support the claim, the authors must either prove attainability of Eq. (3) for n=3 or rephrase the theorem as a comparison of upper bounds.","section":"Section II-B and Theorem 1 n=3 branch"},{"comment":"The comparison is made against the Lo-Popescu protocol, which is optimal for the probability of distilling a Bell state, but the task here is to maximize CHSH. It is not shown that the Lo-Popescu protocol is optimal for CHSH, nor is it justified that the failure state is product (which is assumed in the V_ED formula). Without an optimality proof for CHSH, the claim that nonlocality distillation 'outperforms the optimal entanglement distillation protocol' is not established. A counterexample or a proof that any successful Bell-distillation protocol has failure product and is CHSH-optimal is needed.","section":"Section II-A and Theorem 1"},{"comment":"The proof bounds ||N^2||, which yields an upper bound on the CHSH value, but the theorem claims that the optimal nonlocality distillation protocol 'can attain' a higher CHSH value for n=2. The proof does not explicitly show that the bound is attained for n=2; it only observes that the bound equals the n=1 value. For n=3, the text admits the bound is not known to be tight, so the theorem's statement for n=3 is only an upper bound. Additionally, the comparison to 'optimal entanglement distillation' for mixed states is missing entirely: no mixed-state V_ED is defined or computed. The n=2 L̃ bound also contains a duplicated sum (the same four C terms appear twice with different coefficients), which appears to be a typo and obscures the derivation.","section":"Section III, proof of Theorem 2"}],"minor_comments":[{"comment":"The notation p_succ = min_{1≤r≤2} (2/r)(λ_{2−r+1}+...+λ_8) is unclear for general n; the upper limit of the sum should depend on the dimension of the state. Please clarify the formula and specify the Schmidt dimension for each n.","section":"Eq. (2)"},{"comment":"There is a punctuation error: 'where, The CHSH value is then given by' should be a new sentence. Also, the derivation of the CHSH value for ρ would benefit from explicitly writing the four correlators and the resulting expression.","section":"Section II-C"},{"comment":"The bound ||L1|| ≤ 8|a−b| + 24|c−b| + 8|d−b| is stated without derivation. Please provide the eigenvalue or norm argument, or cite a standard result, so the reader can verify the step.","section":"Section III, Theorem 2 proof"},{"comment":"The intervals stated in the text are inconsistent: for n=3 the proof says the inequality holds for p∈[0.5,0.746] and then concludes it proves the result for p∈[0.746,0.904]. The figure labels should match the text.","section":"Figure 1 and proof of Theorem 1"},{"comment":"The resource comparison is based on a specific circuit implementation and a particular sparse-state-preparation algorithm. Please state clearly that these are implementation-dependent estimates, not general complexity-theoretic claims.","section":"Section III-A"}],"recommendation":"major_revision","confidential_remarks":"The n=2 pure-state comparison may be salvageable after correcting the algebraic errors and clarifying the optimality of the entanglement-distillation side. The n=3 branch, however, currently rests on an unproven attainability statement that the authors themselves flag. The mixed-state theorem also lacks a clear comparison baseline. I recommend major revision, with the requirement that the V_ED formulas be recomputed, the n=3 attainability gap be resolved, and the optimality assumptions be justified."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my read. The paper asks the right question: for a few copies of a bipartite pure state, is it better to distill entanglement first or just perform collective CHSH measurements? The claim that nonlocality distillation wins for n=2 and 3 is plausible and worth knowing. I think the n=2 case may survive contact with a careful proof, and the mixed-state bound in Eq. (6) is a useful addition. The resource comparison is rough but honest. The soft spots are real. The n=3 branch of Theorem 1 depends on Eq. (3) being an attainable CHSH value, but the authors themselves say the measurements have only been verified numerically. That is an upper bound, not an attainment claim, so the 'can attain' is unsupported. This is the main problem. Theorem 2 says 'optimal' but the proof only bounds the norm of a specific operator; it does not show that the bound is achieved or optimal. The proof details also have inconsistencies: the n=3 interval in the proof does not match the theorem statement, and a careful reader will want to check the displayed inequality for n=2 at p=0.5—I suspect a typo, but as printed it gives an impossible CHSH value. The reader's note about Eq. (2) at p=1/2 is a misreading: that state is product, not a Bell state, so zero there is not necessarily wrong. But the formula itself has issues at p=1, where it gives a success probability above one. So the technical core needs serious cleanup. If the n=2 case is correct after cleanup, the paper is a solid finite-copy comparison. The n=3 claim needs either a proof of attainability or a softening to 'the best known nonlocality-distillation protocol achieves...' which would still be new. I would send it to peer review, because the question is significant and a referee can help separate typos from actual contributions. As written, I wouldn't cite it yet. Bring it to reading group if people want to untangle the proof.","headline":"Plausible finite-copy comparison, but the proof has load-bearing gaps—especially the unproven n=3 attainability claim.","tokens_in":648,"tokens_out":596,"would_cite":false,"duration_ms":85702,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for two or three copies of a weakly entangled state, directly distilling nonlocal correlations yields a higher CHSH violation than first distilling a Bell state, even though the latter uses communication.","keywords":["nonlocality distillation","entanglement distillation","CHSH inequality","collective measurements","quantum resource estimation","sparse state preparation","device-independent QKD","small-copy regime"],"falsifier":"Explicitly construct or find the measurements that achieve the n=3 collective CHSH bound and demonstrate a CHSH value above V_ED at some p in [0.746,0.904]. If no such measurements exist—that is, the bound is not attainable—the n=3 claim fails. A direct experiment on two copies at p=0.7 would test the n=2 claim: the measured CHSH should exceed V_ED for p in [0.5,0.85].","tokens_in":8346,"feed_emoji":"⚛️","tokens_out":7766,"duration_ms":73851,"temperature":0.7,"pith_summary":"This paper compares two ways to amplify the nonlocality of a small number of shared partially entangled states. It establishes that for two or three copies of a specific two-qubit pure state, a protocol that directly maximizes the CHSH violation via collective measurements can achieve a higher CHSH value than the optimal entanglement-distillation protocol, which first concentrates a Bell state and then measures it—even though entanglement distillation is allowed one-way communication. The advantage is shown for explicit parameter ranges and disappears by four copies, where entanglement distillation is always better. The comparison extends to a family of mixed states, and a resource estimate indicates the direct nonlocality approach also needs fewer qubits, gates, and runtime.","feed_headline":"Nonlocality distillation beats entanglement distillation for 2–3 copies","feed_subtitle":"Direct collective CHSH measurements outperform first concentrating a Bell state, even with one-way communication allowed.","key_machinery":"The comparison is driven by two quantities: V_ED, the CHSH value after optimal entanglement distillation, which equals 2√2 times the Bell-state success probability plus 2 times the failure probability; and V_ND, the best known CHSH value from collective measurements on n copies, computed from the coefficients λ_i of the n-copy bipartite decomposition via the formula V_ND = 2 Σ_{n=1}^{⌊d/2⌋} √( (λ_{2n−1}^2 + λ_{2n}^2)^2 + 4λ_{2n}^2 λ_{2n−1}^2 ). For mixed states, the proof bounds the operator norm of the effective measurement N by writing N^2 = 4 I⊗K̃ + (2A)⊗L̃ and bounding K̃ and L̃, yielding ∥N∥ ≤ 2√(1+(1−2p)^2) for n=1,2 and a higher-degree polynomial bound for n=3.","core_discovery":"For the state |Ψ⟩ = √p|ψ⟩ + √(1−p)|ϕ⟩ (a superposition of two maximally entangled states), the paper proves that the best known nonlocality-distillation value for n=2 and n=3 copies exceeds the CHSH value of optimal entanglement distillation in the intervals p∈[0.5,0.85] and p∈[0.746,0.904]. The entanglement value is a weighted average: with success probability p_succ the protocol yields a Bell state with CHSH 2√2, otherwise a product state with CHSH 2. The nonlocality value comes from a collective-measurement bound expressed in terms of the coefficients of the n-copy bipartite decomposition. For the mixed state ρ = p|ψ⟩⟨ψ| + (1−p)|ϕ⟩⟨ϕ|, the paper proves the optimal nonlocality-distillation","pith_inferences":["If the n=3 nonlocality-distillation bound is proven tight with explicit measurements (currently only numerical), the advantage window might extend or shift; until then, the n=3 branch of the theorem rests on an unverified attainment assumption.","For device-independent QKD, higher CHSH values from nonlocality distillation could improve key rates or tolerable noise, but only if the distilled correlations have the right secrecy structure—nonlocal correlations do not automatically yield secrecy.","The resource estimate suggests a concrete experimental test: with two copies of |Ψ⟩, directly performing the collective CHSH measurements should outperform the full distillation-plus-measurement circuit on near-term hardware, provided a fair accounting of communication overhead.","The result points to a general principle: when the target is a Bell-inequality violation, the optimal distillation strategy should be optimized for that observable, not for Bell-state fidelity; entanglement concentration is a different goal."],"forward_implications":["For n=2 and n=3 copies of the pure state, nonlocality distillation achieves a higher CHSH value than optimal entanglement distillation for concrete parameter ranges (p∈[0.5,0.85] and p∈[0.746,0.904] respectively).","At n=4 copies, the comparison flips: entanglement distillation is always better, so the advantage is a small-copy phenomenon.","For mixed states, the optimal nonlocality distillation protocol for two copies attains the same CHSH value as for the corresponding pure state, so the pure-state advantage carries over to mixed states with the same form.","Nonlocality distillation requires fewer quantum resources than entanglement distillation in the n=2 pure-state case—about 3 fewer logical qubits, roughly 8 times less logical depth, and about 3 times fewer T states according to the resource estimate.","The advantage exists even though entanglement distillation uses one-way communication; the communicated bit carries no input information, so the comparison isolates the value of the distillation strategy itself."],"fun_headline_variants":["Nonlocality distillation wins for 2–3 copies","CHSH boost: nonlocality beats entanglement distillation","Fewer copies, more nonlocality: new distillation edge","Nonlocality distillation surpasses entanglement for few copies"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The n=3 pure-state advantage assumes the collective-measurement CHSH bound in Eq. (3) is attainable, but the paper states the corresponding measurements have only been verified numerically; if the bound is merely an upper bound, the n=3 branch of Theorem 1 is not established.","fun_headline_variants_meta":{"raw":{"variants":["Nonlocality distillation wins for 2–3 copies","CHSH boost: nonlocality beats entanglement distillation","Fewer copies, more nonlocality: new distillation edge","Nonlocality distillation surpasses entanglement for few copies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000135,"raw_usage":{"total_tokens":930,"prompt_tokens":644,"completion_tokens":286,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":388,"completion_tokens_details":{"reasoning_tokens":220}},"tokens_in":388,"tokens_out":286,"duration_ms":3274,"temperature":1.0,"reasoning_tokens":220,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T19:46:20.645405+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Explicitly construct or find the measurements that achieve the n=3 collective CHSH bound and demonstrate a CHSH value above V_ED at some p in [0.746,0.904]. If no such measurements exist—that is, the bound is not attainable—the n=3 claim fails. A direct experiment on two copies at p=0.7 would test the n=2 claim: the measured CHSH should exceed V_ED for p in [0.5,0.85].","supporting_citations":[],"review_version":1}