arxiv: 2604.01197 · v3 · submitted 2026-04-01 · 🪐 quant-ph · cond-mat.stat-mech· cs.CC· cs.LG

Recognition: 2 theorem links

· Lean Theorem

Learning and Generating Mixed States Prepared by Shallow Channel Circuits

Fangjun Hu , Christian Kokail , Milan Kornja\v{c}a , Pedro L. S. Lopes , Weiyuan Gong , Sheng-Tao Wang , Xun Gao , Stefan Ostermann

Authors on Pith no claims yet

Pith reviewed 2026-05-14 22:24 UTC · model grok-4.3

classification 🪐 quant-ph cond-mat.stat-mechcs.CCcs.LG

keywords mixed quantum statestrivial phaseshallow channel circuitsquantum state learningquantum generative modelslocal reversibilitytrace distancediffusion models

0 comments

The pith

Any mixed state in the trivial phase can be learned and approximately generated by a shallow local channel circuit from measurements alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper shows that mixed quantum states belonging to the trivial phase—those for which a shallow preparation channel circuit exists that preserves local reversibility—can be efficiently recovered using only copies of the state and measurement access. The resulting algorithm returns a shallow local channel circuit whose output state is close to the unknown state in trace distance. Resource costs remain polynomial or quasi-polynomial in the number of qubits whenever the underlying depth is constant or polylogarithmic. The work supplies a structural basis for quantum generative models built from shallow channel circuits and carries over to an efficient classical diffusion-model algorithm with only polynomial overhead.

Core claim

Any mixed state in the trivial phase, defined by the existence of a shallow preparation channel circuit that preserves local reversibility throughout, can be learned from measurement data alone. Given copies of an unknown state from this class, the algorithm outputs a shallow local channel circuit that generates an approximation to the state in trace distance, with sample complexity and runtime polynomial or quasi-polynomial in the number of qubits for constant or polylogarithmic depth and gate locality.

What carries the argument

The trivial phase of mixed states, defined by the existence of a shallow preparation channel circuit that preserves local reversibility at every step.

Load-bearing premise

The unknown mixed state belongs to the trivial phase, so a shallow preparation channel circuit preserving local reversibility exists.

What would settle it

A concrete mixed state that admits a shallow reversible channel-circuit preparation but for which the algorithm returns no circuit whose generated state lies within the claimed trace-distance bound.

Figures

Figures reproduced from arXiv: 2604.01197 by Christian Kokail, Fangjun Hu, Milan Kornja\v{c}a, Pedro L. S. Lopes, Sheng-Tao Wang, Stefan Ostermann, Weiyuan Gong, Xun Gao.

**Figure 2.** Figure 2: Schematic of the general circuit setting. We consider a channel circuit [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: Learning scheme of a 2D trivial phase mixed state (generalized from Figure 13 of [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Schematic for local reversibility: any channel gate [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 5.** Figure 5: (a) Approximate Markovianity under tripartition Λ = [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗

**Figure 6.** Figure 6: (a) Schematic for lightcone decomposition (Fact [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗

**Figure 7.** Figure 7: Schematic for Lemma 9. Local reversibility still holds for any extra channel C acting on E≤ℓ−1(σ), as long as with Supp(C) ∩ Supp(Eℓ,x) = ∅. 3.5 Local inversion within a phase We are going to end this section with an important concept called local inversion, arising from the local reversibility property. This local inversion technique will be quite useful when we prove approximate Markovianity by construct… view at source ↗

**Figure 8.** Figure 8: Schematic for local inversion (Lemma 10). For any S ⊆ λ, there exists a channel PS acting on S, such that PS(ρ) ≈ BS¯(σ). To construct PS, we first assign an ordering of indices to gates in BS\BS¯ – denoted as {G˜m}m∈[N] (where N = 6 in this schematic) – based on the ascending order of the layer indices ℓ, and then we act the local recovery gates {G˜m}m∈[N] (the orange channel gates) based on the descendin… view at source ↗

**Figure 9.** Figure 9: Schematic for the proof of the existence of local recovery maps (Theorem 11). We show the approximate Markovianity by the explicit construction of the local recovery Ψ = Q1 ◦RC(s)\C ◦ PB (light gray region). The step from (a) to (b) is from the local inversion PB(ρ) ≈ BAC(ρ0). The step from (b) to (c) is from the reset channel property RC(s) ◦ BC = RC(s) for Supp(BC) ⊆ C(s). The step from (c) to (d) is fro… view at source ↗

**Figure 10.** Figure 10: Schematic of the partition used in the proof of the existence of the local extension map [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗

**Figure 11.** Figure 11: Schematic for the proof of the existence of local extension maps (Theorem 13). We show the local extendibility by the explicit construction of the local extension TrDE ◦ Φ where Φ = Q1 ◦ RF\CD ◦ PBE. Yellow regions represent the regions consistent with ρ; white regions represent the regions being in |0⟩⟨0|; pink regions represent the edge of backward lightcone for ACD (see Eq. (9)). The step from (a) to (… view at source ↗

**Figure 12.** Figure 12: Details of the covering scheme for the case [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗

**Figure 13.** Figure 13: For a trivial phase mixed state ρ, if there is a local extension map ΦBE→BC learned on the local state ρAinBCDE with high accuracy, then ΦBE→BC is also a local extension map for global state ρABCDE with high accuracy. measurements is finite, leading to statistical estimation error. Such tomography error directly affects the quality of the learned extension map and must be carefully incorporated into the o… view at source ↗

**Figure 14.** Figure 14: Reorganization trick that transfers from the original circuit [PITH_FULL_IMAGE:figures/full_fig_p041_14.png] view at source ↗

read the original abstract

Learning quantum states from measurement data is a central problem in quantum information and computational complexity. In this work, we study the problem of learning to generate mixed states on a finite-dimensional lattice. Motivated by recent developments in mixed state phases of matter, we focus on arbitrary states in the trivial phase. A state belongs to the trivial phase if there exists a shallow preparation channel circuit under which local reversibility is preserved throughout the preparation. We prove that any mixed state in this class can be efficiently learned from measurement access alone. Specifically, given copies of an unknown trivial phase mixed state, our algorithm outputs a shallow local channel circuit that approximately generates this state in trace distance. The sample complexity and runtime are polynomial (or quasi-polynomial) in the number of qubits, assuming constant (or polylogarithmic) circuit depth and gate locality. Importantly, the learner is not given the original preparation circuit and relies only on its existence. Our results provide a structural foundation for quantum generative models based on shallow channel circuits. In the classical limit, our framework also inspires an efficient algorithm for classical diffusion models using only a polynomial overhead of training and generation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper shows that mixed states in the trivial phase can be learned efficiently from copies alone and output as shallow channel circuits, with poly or quasi-poly sample cost depending on depth.

read the letter

The core result is a learning algorithm that, given copies of an unknown mixed state prepared by a shallow channel circuit in the trivial phase, recovers a similar shallow circuit that approximates the state in trace distance. The sample and runtime bounds are polynomial for constant depth and quasi-polynomial for polylog depth, and the learner does not need the original circuit, only that one exists. This extends earlier pure-state learning results to mixed states prepared by channels rather than unitaries, and the classical limit sketched for diffusion models is a clean side observation. The argument is conditional on the state belonging to the defined trivial phase, which is stated up front and avoids overclaiming universality. The main limitation is that the depth and locality restrictions are baked into the guarantee; states outside this class are not addressed, and the concrete constants in the sample complexity are not spelled out in the abstract. The proof strategy appears to use local reversibility to build a measurement-based reconstruction, but without the full derivation it is hard to judge how tight the bounds end up. Overall the work is a solid structural contribution for people working on quantum generative models and mixed-state phases. It is worth sending to peer review because the claim is clearly stated, the setting is well-motivated, and the result is falsifiable once the details are checked.

Referee Report

0 major / 3 minor

Summary. The paper claims that any mixed state in the trivial phase—defined as one for which there exists a shallow preparation channel circuit preserving local reversibility throughout—can be efficiently learned from measurement access to copies alone. The algorithm outputs a shallow local channel circuit that approximates the unknown state to trace distance, with sample complexity and runtime polynomial (or quasi-polynomial) in system size for constant (or polylog) depth and gate locality. The learner does not receive the original circuit and the result is conditional on membership in this class; a classical limit for diffusion models is also sketched.

Significance. If the central learning theorem holds, the work supplies a constructive structural result linking mixed-state phases of matter to efficient generative modeling via shallow channel circuits. It gives an explicit measurement-based learner with controlled complexity and suggests a polynomial-overhead classical analogue, which could inform both quantum state tomography and classical generative modeling literature.

minor comments (3)

[Abstract] The abstract and introduction should state the precise dependence of the sample complexity on depth d and locality k (e.g., whether the bound is poly(n,2^d) or involves additional factors) rather than the current high-level “polynomial or quasi-polynomial” phrasing.
[§2] Clarify the precise notion of “local reversibility preserved throughout the preparation” when the circuit is applied to a mixed state; a short paragraph or diagram in §2 would help readers verify that the definition matches the learner’s assumptions.
[Discussion] The classical diffusion-model corollary is stated only in the abstract; a brief self-contained paragraph in the discussion section would make the claimed polynomial overhead explicit and allow direct comparison with existing classical results.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our work and for recommending minor revision. The provided report contains no enumerated major comments, so we have no specific points requiring point-by-point rebuttal. We will incorporate any minor editorial suggestions in the revised manuscript.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The paper defines the trivial phase by the existence of a shallow preparation channel circuit preserving local reversibility, then proves that states in this class admit an efficient measurement-based learner outputting an approximating circuit. This is a standard conditional result: the class membership is an external assumption on the unknown state, and the algorithm constructs the circuit from data without reducing to fitted parameters, self-referential equations, or load-bearing self-citations. The abstract and claims exhibit no self-definitional steps, fitted-input predictions, or ansatz smuggling. The result is independent of the specific circuit and relies on measurement access alone.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that the state is in the trivial phase (existence of a shallow reversible preparation circuit) and on standard quantum information axioms such as the ability to perform local measurements and the definition of trace distance.

axioms (1)

domain assumption Existence of a shallow preparation channel circuit preserving local reversibility for states in the trivial phase
This defines the class of states for which the learning algorithm is claimed to work.

pith-pipeline@v0.9.0 · 5533 in / 1247 out tokens · 44992 ms · 2026-05-14T22:24:30.480788+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/ArithmeticFromLogic.lean; IndisputableMonolith/Cost/FunctionalEquation.lean reality_from_one_distinction; washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

A state belongs to the trivial phase if there exists a shallow preparation channel circuit under which local reversibility is preserved throughout the preparation... local recovery map Ψ_{B→BC}... local extension map Φ_{BE→BC}
IndisputableMonolith/Foundation/DimensionForcing.lean; IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking; D3_admits_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We prove that any mixed state in this class can be efficiently learned... (k+1)-layer channel circuit W

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Concurrence of Symmetry Breaking and Nonlocality Phase Transitions in Diffusion Models
cs.LG 2026-05 unverdicted novelty 7.0

Symmetry breaking and nonlocality phase transitions occur nearly simultaneously during diffusion model generation in modern transformers.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages · cited by 1 Pith paper · 1 internal anchor

[1]

Anderson.Reverse-time diffusion equation models

[And82] Brian D.O. Anderson.Reverse-time diffusion equation models. In:Stochastic Processes and their Applications12.3 (1982), pp. 313–326. [Bak+25] Ainesh Bakshi, John Bostanci, William Kretschmer, Zeph Landau, Jerry Li, Allen Liu, Ryan O’Donnell, and Ewin Tang.Learning the closest product state. In:Proceedings of the 57th Annual ACM Symposium on Theory ...

work page arXiv 1982
[2]

In: arXiv:2504.02208 [quant-ph](2025)

[CR25] Chi-Fang Chen and Cambyse Rouz´ e.Quantum Gibbs states are locally Markovian. In: arXiv:2504.02208 [quant-ph](2025). [Fef+25] Bill Fefferman, Soumik Ghosh, Makrand Sinha, and Henry Yuen.The hardness of learning quantum circuits and its cryptographic applications. In:arXiv:2504.15343 [quant-ph](2025). [FR15] Omar Fawzi and Renato Renner.Quantum Cond...

work page arXiv 2025
[3]

Duan.Quantum Supremacy for Simulating a Translation-Invariant Ising Spin Model

[GWD17] Xun Gao, Sheng-Tao Wang, and L.-M. Duan.Quantum Supremacy for Simulating a Translation-Invariant Ising Spin Model. In:Physical Review Letters118.4 (2017), p. 040502. [Haa11] Jeongwan Haah.Local stabilizer codes in three dimensions without string logical oper- ators. In:Physical Review A83.4 (2011), p. 042330. [Haa16] Jeongwan Haah.An Invariant of ...

work page 2017
[4]

Local Diffusion Models and Phases of Data Distributions

[HKP20] Hsin-Yuan Huang, Richard Kueng, and John Preskill.Predicting many properties of a quantum system from very few measurements. In:Nature Physics16.10 (2020), pp. 1050–1057. 33 [Hu+25] Fangjun Hu, Guangkuo Liu, Yifan Zhang, and Xun Gao.Local Diffusion Models and Phases of Data Distributions. In:arXiv:2508.06614 [cs.LG](2025). [Hua+24] Hsin-Yuan Huang...

work page internal anchor Pith review Pith/arXiv arXiv 2020
[5]

In:arXiv:2509.15204 [quant-ph](2025)

[MKS25] Ruochen Ma, Vedika Khemani, and Shengqi Sang.Circuit-based characterization of finite-temperature quantum phases and self-correcting quantum memory. In:arXiv:2509.15204 [quant-ph](2025). [MT25] Ruochen Ma and Alex Turzillo.Symmetry-Protected Topological Phases of Mixed States in the Doubled Space. In:PRX Quantum6.1 (2025), p. 010348. [MW23] Ruoche...

work page arXiv 2025
[6]

In:arXiv:2511.15806 [quant-ph](2025)

[Pel+25] Angelos Pelecanos, Jack Spilecki, Ewin Tang, and John Wright.Mixed state tomography reduces to pure state tomography. In:arXiv:2511.15806 [quant-ph](2025). [San+25] Shengqi Sang, Leonardo A. Lessa, Roger S. K. Mong, Tarun Grover, Chong Wang, and Timothy H. Hsieh.Mixed-state phases from local reversibility. In:arXiv:2507.02292 [quant-ph](2025). [S...

work page arXiv 2025
[7]

Yao, and Hsin-Yuan Huang.Hardness of recognizing phases of matter

[Sch+25] Thomas Schuster, Dominik Kufel, Norman Y. Yao, and Hsin-Yuan Huang.Hardness of recognizing phases of matter. In:arXiv:2510.08503 [quant-ph](2025). [SE19] Yang Song and Stefano Ermon.Generative Modeling by Estimating Gradients of the Data Distribution. In:Proceedings of the 33rd International Conference on Neural In- formation Processing Systems. ...

work page arXiv 2025
[8]

Hsieh.Mixed-State Quantum Phases: Renormalization and Quantum Error Correction

[SZH24] Shengqi Sang, Yijian Zou, and Timothy H. Hsieh.Mixed-State Quantum Phases: Renormalization and Quantum Error Correction. In:Physical Review X14.3 (2024), p. 031044. [Vot+25] Matteo Votto, Marko Ljubotina, C´ ecilia Lancien, J. Ignacio Cirac, Peter Zoller, Maksym Serbyn, Lorenzo Piroli, and Benoˆ ıt Vermersch.Learning mixed quantum states in large-...

work page arXiv 2024
[9]

Wen.Vacuum degeneracy of chiral spin states in compactified space

[Wen89] X.-G. Wen.Vacuum degeneracy of chiral spin states in compactified space. In:Physical Review B40.10 (1989), pp. 7387–7390. [Wen90] X.-G. Wen.Topological Orders in Rigid States. In:International Journal of Modern Physics B04.02 (1990), pp. 239–271. [YSL25] Tai-Hsuan Yang, Bowen Shi, and Jong Yeon Lee.Topological Mixed States: Phases of Matter from A...

work page arXiv 1989
[10]

In:Physical Review Letters132.10 (2024), p

[Zha+24] Bingzhi Zhang, Peng Xu, Xiaohui Chen, and Quntao Zhuang.Generative Quantum Machine Learning via Denoising Diffusion Probabilistic Models. In:Physical Review Letters132.10 (2024), p. 100602. A Proofs of Backward Lightcone Decompositions In this appendix, we provide the proof of backward lightcone decompositions (Fact 3 and Fact

work page 2024