aiqbench analysis
This research establishes a rigorous framework for certifying temporal reports in time-series forecasting by distinguishing between the identifiability of delay indicators and their actual functional use within model computations. By deriving exact recovery identities for Gaussian point delays, the study reveals significant gaps where reported correlations do not align with true computational dependencies in architectures like TCN and N-HiTS. These findings are critical for validating the reliability of AI-generated forecasts that rely on auxiliary historical data or leading indicators.
From Recoverability to Functional Use: Certifying Temporal Reports in Time-Series Forecasting
arXiv:2608.10433v3 Announce Type: replace Abstract: Models increasingly accompany time-series forecasts with temporal reports---delays, leading indicators, or selected history---yet a correct report need not describe the computation that produced the forecast. We formalize this as a three-stage certification problem: \emph{recoverability} of the target from the realized trajectory, \emph{correctness} of the model's report, and \emph{functional use} of the reported history. For Gaussian point delays, we derive an exact finite-sample recovery--substitutability identity: the same realized shift geometry yields structural evidence at scale $n\eta_n$ but normalized proxy-prediction cost at scale $\eta_n$. Thus a delay can be decisively identifiable while a correlated alternative remains near-oracle. We then audit intentionally unconstrained TCN and N-HiTS forecasters only on trajectories that are recoverable, correctly reported, and near-oracle. Even there, Jacobian and in-distribution conditional-replacement response peaks lie $9.3$--$13.5$ time steps from the reported delay. Finally, a report-conditioned no-bypass factorization provides a sufficient access certificate; architecture-matched post-hoc routing and gate-destruction controls show that the alignment change is attributable to report-coordinate access. The framework distinguishes evidence that a temporal statement is identifiable and correct from evidence that the forecast computation actually depends on it.
Abstract:Models increasingly accompany time-series forecasts with temporal reports---delays, leading indicators, or selected history---yet a correct report need not describe the computation that produced the forecast. We formalize this as a three-stage certification problem: \emph{recoverability} of the target from the realized trajectory, \emph{correctness} of the model's report, and \emph{functional use} of the reported history. For Gaussian point delays, we derive an exact finite-sample recovery--substitutability identity: the same realized shift geometry yields structural evidence at scale $n\eta_n$ but normalized proxy-prediction cost at scale $\eta_n$. Thus a delay can be decisively identifiable while a correlated alternative remains near-oracle. We then audit intentionally unconstrained TCN and N-HiTS forecasters only on trajectories that are recoverable, correctly reported, and near-oracle. Even there, Jacobian and in-distribution conditional-replacement response peaks lie $9.3$--$13.5$ time steps from the reported delay. Finally, a report-conditioned no-bypass factorization provides a sufficient access certificate; architecture-matched post-hoc routing and gate-destruction controls show that the alignment change is attributable to report-coordinate access. The framework distinguishes evidence that a temporal statement is identifiable and correct from evidence that the forecast computation actually depends on it.
Machine Learning (cs.LG)
arXiv:2608.10433 [cs.LG]
(or arXiv:2608.10433v3 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.10433 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 270 words
