QMA, or quantum Merlin-Arthur, is the complexity class of promise problems whose yes instances have a polynomial-size quantum certificate accepted
with probability at least 
by a polynomial time quantum verifier, while every
certificate for a no instance is accepted with probability at most 
. The verifier is given one quantum certificate, which may
contain entanglement among its qubits.
In QMA(2), the verifier receives two certificates promised to be unentangled with one another. This restriction applies to the admissible proofs, not to the verifier's subsequent joint computation. Clearly QMA is contained in QMA(2), since the second certificate can be ignored.
Bostanci et al. (2026) constructed a quantum oracle relative to which the containment is strict. This is an oracle separation and does not prove the unrelativized classes unequal. They also resolved the no-disentanglers conjecture, giving an exponential lower bound on the input size of channels approximating the set of separable output states. GPT-5.6 Sol supplied a central construction that the authors developed and checked. Independent external review had not been reported as of Sep. 7, 2026.
