The Signal
Quantum channel paper sets new average-case testing bounds
A paper published in Quantum sets a lower bound for worst-case quantum channel testing and defines an average-case alternative. It gives query bounds for general channels and dimension-independent bounds for three channel classes under that alternative.

A paper published in Quantum on Nov. 18 sets a lower bound for worst-case quantum channel testing and gives query bounds using a new average-case distance. The study by Gregory Rosenthal, Hugo Aaronson, Sathyawageeswar Subramanian, Animesh Datta and Tom Gur also establishes dimension-independent query bounds for testing three specific classes of channels under that distance.
A floor for worst-case testing
The paper reports that testing whether a black-box channel matches a fixed target, or is at least ε away in diamond distance, requires Ω(√din/ε) queries. The lower bound applies even to algorithms with ancillae, coherent queries and adaptive choices, and assumes dout is at least 2. The paper frames this as a limit on testing against the diamond norm, which measures the greatest separation over possible inputs rather than typical behavior.
For pure-state replacement channels, which discard the input and return a fixed pure state, the authors prove a stronger lower bound: Ω(din/ε) queries in the same strongest query model. Their results concern theoretical query complexity, not measurements on a physical quantum processor.
A different distance
To study average-case behavior, the authors introduce the average-case imitation diamond norm, or ACID norm. It measures the trace-norm difference between the channels’ Choi operators, formed by applying each channel to one half of a maximally entangled state. By contrast, the diamond norm maximizes over inputs; the paper notes that diamond distance is at least as large as ACID distance.
The distinction matters to the bounds the paper proves. In the weakest query model, with no ancillae, coherence or adaptivity, testing an arbitrary fixed channel in ACID distance takes at most Õ(dindout3/2/ε2) queries. For comparison, the paper gives an upper bound of Õ(din2dout3/2/ε2) for testing an arbitrary fixed channel in diamond distance in that model.
Bounds for particular channels
For erasure channels, unitary channels and pure-state replacement channels, the authors give a dimension-independent ACID-distance bound of O(1/ε2) queries. The stated test accepts with probability 1 when the channel equals the target and with probability at most 1/2 when its ACID distance from the target is at least ε. The result is limited to those channel types; it is not a dimension-independent bound for arbitrary channels.
The paper also establishes tight bounds for channel tomography in ACID distance. In its comparison of query models, it gives a coherent tomography upper bound of O(din2dout2/ε2) and, for output dimension of at least 4, an incoherent lower bound of Ω(din3dout3/ε2).
Publication history
The work first appeared on arXiv on Sept. 19, 2024, and the listed version was revised on Nov. 13, 2026. The journal publication identifies it as Quantum volume 10, article 2276.
Sources
Related stories
- Mele and Bittel match channel tomography’s dimension scaling
- Quantum paper maps a matrix-based route to channel identification
- Yang's Counterexample Defeats a Quantum-Channel Optimality Conjecture
- Three Researchers Derive a Unique Quantum Bayes Rule
- Researchers define states whose entanglement dimension resists change