← All problems

Marton’s conjecture on broadcast-channel capacity

For every finite two-receiver discrete memoryless broadcast channel pY,Z∣Xp_{Y,Z\mid X}, the capacity region equals the full Marton inner bound: C(pY,Z∣X)=RMarton(pY,Z∣X)\mathcal{C}(p_{Y,Z\mid X})=\mathcal{R}_{\mathrm{Marton}}(p_{Y,Z\mid X}).

References

Source: arXiv

Progress summary

Refreshed
Open · solution reported

A new preprint claims to disprove Marton’s conjecture, but the result has not yet been independently checked.

Marton’s conjecture asks whether the full Marton coding method gives the capacity of every two-receiver broadcast channel; it has guided work on the problem since 1979.

Known results

  • Marton’s region remains the largest known general achievable region, while the general capacity region is unknown (Cover, 1998).
  • Körner and Marton determined capacity for the more-capable class (1979).
  • The no-WW version of Marton’s region was shown non-optimal in examples, but this does not address the full region (Geng, Nair, et al., 2010).

August 2026 claimed counterexample

Yanxiao Liu and Mian Huang claim a finite discrete memoryless broadcast channel for which the complete one-letter Marton region is strictly smaller than capacity, disproving the conjecture. The argument is presented in a preprint and has no independent verification, correction, or retraction in the supplied evidence. The authors report assistance from GPT-5.6 Sol, Claude Fable 5, and Opus 5.

Current status (as of August 2026): Marton’s conjecture remains unverified as a theorem, but a preprint claims a counterexample showing that the full Marton inner bound is not generally optimal.

Sources

Solutions 0

No solutions have been posted yet.