Overview
Research has identified two distinct counterexamples to the Markovity Conjecture, originally proposed by Gohari, Liu, and Nair (ISIT 2025). This conjecture posited a specific structural property for the optimizers of the dual functional associated with Marton's inner bound. If the conjecture had held true, it would have significantly simplified the evaluation process of Marton's inner bound.
Research Context
The Markovity Conjecture relates to the dual functional that arises in the context of Marton's inner bound. Marton's inner bound is a key component in information theory, particularly in characterizing the capacity region of multi-user communication channels. The conjecture specifically concerned the structural characteristics of the optimizers of this dual functional. A true Markovity Conjecture would have streamlined the computational complexity involved in assessing Marton's inner bound, implying a simpler, more predictable structure for optimal solutions.
Approach
The researchers developed two specific constructions to challenge the Markovity Conjecture. Both constructions share common characteristics:
- They are based on ternary-input broadcast channels.
- They exhibit strictly positive transition probabilities.
- They utilize an identical nonrectangular $2\times2$ auxiliary structure.
For each construction, a non-Markov configuration was explicitly generated. The objective value of these non-Markov constructions was then rigorously compared against the maximum achievable objective value from any construction adhering to the conjectured Markov structure. The development of these examples was facilitated by the assistance of GPT-5.6 Sol.
Findings
The study found that in both developed scenarios, the explicit non-Markov constructions yielded an objective value that was rigorously larger than any objective value attainable by constructions that satisfied the Markov structure as conjectured. This direct comparison demonstrates that the non-Markov structures outperform their Markovian counterparts in these specific cases.
These findings collectively serve to disprove the Markovity Conjecture. The existence of even one such counterexample is sufficient to invalidate a universal conjecture regarding structural properties.
Why This Matters
The Markovity Conjecture, if valid, would have streamlined computations related to Marton's inner bound, which is relevant for the study of communication channel capacity. The disproof of this conjecture indicates that the optimization landscape for Marton's inner bound is more complex than previously hypothesized, and that simpler structural assumptions may not hold true in all cases. This outcome implies that approaches to evaluating Marton's inner bound will need to consider more general, non-Markovian structures, which may involve increased computational complexity or require alternative simplification strategies.