A deblurring trick from 1970s astronomy is beating the leading quantum error mitigation method on benchmark circuits. The catch: the leading method has been beaten before, and nothing shipped.
image from grok
Researchers at Florida State University adapted the 1970s Richardson-Lucy astronomy deblurring algorithm into HAMMR-L, a quantum error mitigation technique that models hardware noise as a convolution effect across Hamming distance and inverts it to surface correct measurement outcomes. On IBM quantum hardware, HAMMR-L outperformed the leading QBEEP method in 7 of 10 benchmarks, improving the correct answer from rank ~4 (less than 1% probability) to rank 1 (nearly 8% probability) on a nine-qubit Bernstein-Vazirani circuit after ten thousand shots. The technique remains unreleased and untested on practical circuits with entangled states, echoing a persistent pattern where quantum error mitigation papers fail to transition to production.
A deblurring algorithm from 1970s astronomy is having a second career in quantum computing. HAMMR-L, a technique adapted from the Richardson-Lucy method originally developed to sharpen telescope images, improves the ranking of correct measurement outcomes in quantum circuits by modeling noise as a kind of blur across a state graph and then removing it. The approach, described in a preprint posted to arXiv on March 29, 2026 by five researchers at Florida State University, outperformed the current leading method in seven of ten benchmark datasets on IBM quantum hardware.
The technique works by reframing the quantum measurement problem. When a quantum circuit runs, the output distribution is smeared by hardware noise — the correct answer gets buried under erroneous outcomes that look similar in the circuit's mathematical representation. HAMMR-L treats that smearing as a convolution effect across Hamming distance (the number of bit positions at which two output strings differ) and inverts it. The result: the correct string surfaces rather than staying hidden at rank four with less than 1 percent probability, as the researchers observed on a nine-qubit Bernstein-Vazirani circuit running on IBM Brisbane. After ten thousand shots with deconvolution applied, the correct answer held rank one with nearly 8 percent probability.
The comparison point is QBEEP, a Hamming-distance-based error mitigation method published by researchers at Pacific Northwest National Laboratory (PNNL) at the ISCA 2023 conference. QBEEP remains the standard in this class of techniques. On the same ten-dataset benchmark suite, HAMMR-L achieved a mean rank improvement of plus 3.081 across circuits. QBEEP's mean rank change was minus 0.328 — it made things slightly worse on average.
Bernstein-Vazirani circuits are a useful test case because they have a known hidden string as the answer, which makes it straightforward to check whether a technique is working. The limitation is that the paper explicitly describes them as "not hugely useful" for practical quantum computation — they are a laboratory tool, not a workload. The technique's performance on more complex circuits with entangled states remains untested in this work.
There is also the question of what happened to QBEEP itself. The PNNL team published their method in 2022 with similar claims about improving measurement outcomes. Three years later, the technique has not been adopted by any major quantum computing platform, does not appear in IBM's Qiskit or Google's Cirq libraries, and has not shipped in any vendor's error mitigation stack. The pattern is worth noting: a preprint circulates with benchmark numbers, the quantum press covers it as a development worth watching, and then nothing ships.
HAMMR-L has no code release announced and no vendor adoption. The authors — Jake Scally, Austin Myers, Ryan Carmichael, Phat Tran, and Xiuwen Liu, all at Florida State University — submitted the work as a quant-ph preprint on March 29. It has not been peer reviewed.
This is the recurring cycle in quantum error mitigation: a new technique appears, beats existing methods on a benchmark suite, gets written up as a potential breakthrough, and then disappears into the gap between preprint and product. The underlying problem is real — quantum hardware noise genuinely corrupts measurement outcomes, and any method that can partially reverse that corruption is worth studying. Whether this particular method generalizes beyond Bernstein-Vazirani circuits, survives the move from a nine-qubit IBM Brisbane run to a hundred-qubit system, and can be implemented without requiring ten thousand shots per query are questions the preprint does not answer.
The Richardson-Lucy algorithm itself is not the story. It has been used in astronomy, medical imaging, and electron microscopy for fifty years. What matters is whether it transfers intact to quantum state spaces, whether it scales, and whether anyone will actually run it on real hardware outside of a benchmark. Right now the answer to all three questions is: unclear.
The honest version of this story is that five researchers at Florida State University showed a deconvolution trick from 1970s image processing can improve rankings of correct answers on a narrow class of quantum circuits, on IBM hardware, in a non-peer-reviewed preprint. That is a result worth tracking. It is not yet a product, a standard, or a solved problem.
Story entered the newsroom
Research completed — 0 sources registered. HAMMR-L applies Richardson-Lucy image deconvolution (from astronomy/microscopy) to quantum state graphs in Hamming space. Errors are treated as blur,
Draft (690 words)
Approved for publication
Headline selected: The Right Quantum Answer Was There All Along
Published (690 words)
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Quantum Computing · 14h 37m ago · 2 min read
Quantum Computing · 15h 34m ago · 4 min read
