MIP*=RE: Quantum Entangled Interactive Proofs

• MIP*=RE is a complexity class equality connecting quantum entangled interactive proofs to recursively enumerable languages, extending even to undecidable problems.
• The proof leverages quantum low-degree tests, recursive compression, and PCP composition to enforce rigidity and preserve soundness gaps.
• Key implications include the undecidability of nonlocal game values, a negative resolution of Tsirelson’s problem, and a counterexample to Connes’ embedding conjecture.

The complexity class equality $\mathrm{MIP}^* = \mathrm{RE}$ captures a fundamental leap in our understanding of multiprover interactive proofs (MIP) with quantum entangled provers. Specifically, it asserts that any recursively enumerable language admits a classical verifier that can be convinced of membership via polynomially bounded interaction with entangled quantum provers—a capability that exactly matches the power of Turing machines and includes even undecidable languages. This equivalence, proven by Ji, Natarajan, Vidick, Wright, and Yuen, has far-reaching consequences for computational complexity, quantum information theory, and operator algebras.

1. Background: Classical and Quantum MIP Protocols

The original model of multiprover interactive proofs (MIP) considers a classical verifier interacting with multiple non-communicating provers. Babai, Fortnow, and Lund established that the classical MIP class characterizes the nondeterministic exponential time languages ($\mathrm{NEXP}$), i.e., $\mathrm{MIP} = \mathrm{NEXP}$ (Natarajan et al., 2019). The protocols leverage isolated yet colluding provers, and the verifier's randomness to extract verifiable information.

Quantum variants, permitting the provers to share arbitrary entanglement, define the class $\mathrm{MIP}^*$. Early bounds showed at least $\mathrm{NEXP} \subseteq \mathrm{MIP}^*$; however, the exact characterization of $\mathrm{MIP}^*$ remained unresolved until it was shown that $\mathrm{MIP}^* = \mathrm{RE}$ (Ji et al., 2020). This outcome implies that quantum entanglement dramatically boosts the power of interactive proof systems, subsuming all recursively enumerable languages—including undecidable ones such as the Halting problem.

Key distinctions:

Model	Completeness Class	Prover Resources
Classical MIP	$\mathrm{NEXP}$	Local/shared randomness
Quantum MIP ($\mathrm{MIP}^*$)	$\mathrm{RE}$	Shared entanglement (unbounded)

2. Core Proof Techniques Leading to $\mathrm{MIP}^* = \mathrm{RE}$

The proof of $\mathrm{MIP}^* = \mathrm{RE}$ synthesizes tools from quantum self-testing, PCP theory, algebraic techniques, and recursive protocol compression. The major technical pillars are:

• Quantum Low-Degree Tests: Provably enforce that entangled provers behave as if their answers are evaluations of low-degree polynomials over finite fields, even under entanglement. Strategies passing these tests are close to ideal (up to an error $\delta$ polynomially related to the test soundness $\varepsilon$), ensuring rigidity.

$\operatorname{Pr}[f(x) = g(x)] \leq \frac{d}{q}$

• Recursive Compression Framework: The verifier's workload is exponentially compressed through introspective self-testing and gap-preserving reductions, iteratively encoding the work of a Turing machine into the strategy space of entangled provers. For an original protocol with entangled value $\omega^*(G)$, compressed versions maintain the soundness gap:
  • If $\omega^*(G) = 1$ then $\omega^*(G') = 1$
  • If $\omega^*(G) \leq 1/2$ then $\omega^*(G') \leq 1/2$.
• Classical and Quantum PCP Composition: Classical PCP theorems for succinct encodings, when combined with quantum low-degree testing, allow the encoding and verification of exponentially long proofs through polynomial means, extended to quantum correlations.
• Introspective Protocols: Provers sample most of their questions locally on entangled EPR registers, with the verifier enforcing honest measurement and "hiding" registers via tests rooted in quantum uncertainty—preventing illicit information leakage between provers.

These mechanisms allow the encoding of any Turing machine computation into a nonlocal game whose entangled value $\omega^*(G_M)$ signals whether $M$ halts.

3. Computational and Physical Implications

The identification $\mathrm{MIP}^* = \mathrm{RE}$ yields several sharp consequences:

• Undecidability of Entangled Value Approximation: It is undecidable to determine, for general nonlocal games $G$, whether $\omega^*(G) = 1$ or $\omega^*(G) \leq 1/2$ (Ji et al., 2020). This reduction from the Halting problem is efficient and robust under gap-preserving reductions, establishing that all RE languages (including undecidable problems) reduce to the nonlocal value decision problem (Mančinska et al., 8 May 2025).
• Negative Resolution of Tsirelson's Problem: The set of quantum tensor product correlations ($C_{qa}$) is strictly contained in the set of quantum commuting correlations ($C_{qc}$); i.e., $C_{qa} \subsetneq C_{qc}$. There exist nonlocal games where the commuting operator model achieves value 1, but all finite-dimensional tensor product strategies are bounded away from 1. This answers Tsirelson's problem negatively and demonstrates the non-equivalence of operator models of quantum correlations (Cabello et al., 2023).
• Refutation of Connes' Embedding Conjecture: $C_{qa} \neq C_{qc}$ provides a counterexample to Connes' embedding conjecture, as certain type II$_1$ von Neumann algebras cannot be embedded into the ultrapower of the hyperfinite II$_1$ factor.

4. Expressive Power, Reductions, and Hierarchical Complexity

The scope of $\mathrm{MIP}^*$ was further explored by contrasting constant gap and zero-gap variants (Mousavi et al., 2020). In the promise version, the gapped protocol captures precisely $\mathrm{RE}$. If the gap is removed, the corresponding decision problem rises to the class $\Pi_2^0$ in the arithmetical hierarchy, i.e., languages decidable with a $\forall\exists$ quantified formula in a computable predicate:

$$L = \{x : \forall y \exists z\, R(x, y, z) = 1 \text{ for a computable } R \}$$

Thus, $\mathrm{MIP}^*_0 = \Pi^0_2$ strictly exceeds the expressive power of $\mathrm{RE}$.

Gap-preserving reductions remain integral to these complexity translational arguments. The construction of independent set games, with precise gap maintenance and a new stability theorem for perturbed operator families, establishes MIP*$^*$-completeness for these natural graph-theoretic games. This is in direct contrast to the classical setting, where such problems are in $\mathrm{NP}$ and decidable (Mančinska et al., 8 May 2025).

5. Succinct, Zero-Knowledge, and Robustness Results

Recent developments have extended the power of $\mathrm{MIP}^* = \mathrm{RE}$ protocols to perfect zero-knowledge (PZK) and succinct formulations. Through a series of compression—question reduction, oracularization, answer reduction, and parallel repetition—it is possible to construct two-prover, one-round PZK MIP* protocols for $\mathrm{RE}$ with either:

• Polylogarithmic question size and $O(1)$ answer size, or
• $O(1)$ question size and polylogarithmic answer size

while preserving the zero-knowledge property (Fu et al., 6 Mar 2025). These results exploit:

• The equivalence between two main variants of binary constraint system (BCS) nonlocal games (constraint-constraint and constraint-variable formulations).
• Algebraic analysis (weighted *-algebras and quantum soundness toolkits) to ensure that the quantum and even commuting-operator soundness gap is preserved after conversion.
• Theoretical equivalence between near-perfect strategies across different BCS game types.

This demonstrates that MIP* protocols for even undecidable languages can be made both privacy-preserving and communication-efficient (Mastel et al., 1 Apr 2024).

6. Noise Sensitivity and Limits of Quantum-Entangled Proofs

One critical assumption underlying the immense expressive power of $\mathrm{MIP}^*$ is the availability of unbounded, noiseless entanglement. Recent work has clarified that this advantage collapses with the introduction of noise. If provers share only noisy EPR pairs—even with arbitrarily small constant noise per pair—then the overall class collapses to $\mathrm{NEXP} = \mathrm{MIP}$, i.e., the classical bound (Dong et al., 2023). This separation is tight; perfect (noiseless) entanglement restores the full $\mathrm{RE}$ power. Technical innovations include deterministic positivity testers for large matrices and invariance principles for smooth matrix functionals.

This establishes that the theoretical extremity of $\mathrm{MIP}^*$ is not robust to practical imperfections, imposing severe constraints on experimental realization and device-independent cryptography in the noisy regime.

7. Broader Significance and Open Questions

The equality $\mathrm{MIP}^* = \mathrm{RE}$ and its consequences constitute a major advance in quantum complexity and mathematical physics. The separation $C_{qa} \neq C_{qc}$, refutation of Connes' embedding conjecture, and undecidability of nonlocal game value usher quantum information theory into the field of classical undecidability. Still unresolved is the physical attainability of extreme correlations (those in $C_{qc} \setminus C_{qa}$). Possibilities include:

1. Physical systems cannot realize such correlations at all (hidden finite-dimensionality).
2. They may only arise in contextuality rather than true nonlocality scenarios.
3. They require violation of spacelike separation.
4. They are physically attainable even under spacelike separation, challenging foundational quantum and relativistic principles (Cabello et al., 2023).

Ultimate answers to these open problems will have foundational implications for quantum nonlocality, the theory of operator algebras, and potentially for the structure of spacetime itself.

---

Representative mathematical expressions:

• Entangled value of nonlocal game $G$:

$$\omega^*(G) = \sup_{\text{quantum strategies}} \sum_{x,y,a,b} \pi(x, y) V(a, b|x, y) \langle \psi | A_x^a \otimes B_y^b | \psi \rangle$$

• Gap-preserving reduction $T$:

$$\text{If } \omega^*(G) \geq 1 \Rightarrow \omega^*(T(G)) \geq 1-\epsilon, \quad \omega^*(G) \leq 1-\delta \Rightarrow \omega^*(T(G)) \leq 1-f(\delta)$$

• Hierarchy of correlation sets:

$$C_q \subsetneq C_{qs} \subsetneq C_{qa} \subsetneq C_{qc}$$

• Example from arithmetical hierarchy:

$L = \{ x : \forall y \exists z\, R(x, y, z) = 1 \}$

---

The realization of $\mathrm{MIP}^* = \mathrm{RE}$ recasts interactive proofs with entangled provers as a touchstone for both computational power and the architecture of quantum theory, providing the first natural complexity-theoretic explanation for the separation of physically meaningful quantum correlation models, the undecidability of broad classes of operator-algebraic questions, and the profound power as well as impermanence of quantum interactive proofs in the presence of entanglement.