Implications of Costly or Imperfect Verification for Computational Arbitrage

Investigate the implications of introducing nonzero verification costs or imperfect verification accuracy into the computational arbitrage framework for AI model marketplaces, assessing how these verification frictions alter the analysis and conclusions derived under the assumption of costless, perfect verification.

Background

The paper analyzes arbitrage in AI model markets under the simplifying assumption that solutions can be verified costlessly and perfectly, which enables clear comparisons of cost-to-solution across providers and arbitrage policies.

The authors explicitly note that many questions remain open and flag the need to study what happens when verification is costly or imperfect, a realistic scenario in many domains where verification requires human effort or may yield false positives or negatives.

References

Many theoretical and empirical questions remain open. In this work we assume verification to be costless. An important research avenue is to investigate the implications of costly or imperfect verification~\citep{gdpval}.

— Computational Arbitrage in AI Model Markets  (2603.22404 - Olmedo et al., 23 Mar 2026) in Discussion (Section 7)

Because latency was not experimentally manipulated in our study, future experiments should also test whether this friction functions as a costly signal that improves discernment, or merely as an engineering constraint.

— Beyond "Made with AI": Visualizing Provenance Density to Mitigate the Transparency Penalty  (2609.03460 - Zhang et al., 3 Sep 2026) in Section 5, Limitations and Future Work, subsection “Latency Constraints”