Dynamical correlation functions of extensive charges after global quantum quenches

arXiv:2607.19208 2026 Architecture 1 ideas extracted · analyzed Aug 30, 2026

What the math gives to ML

The paper offers a nontrivial time-shell factorization of multi-time full-counting statistics after quenches with sufficiently weak temporal correlations, especially ballistic transport. Its transferable asset is that a large multi-time dependence can be represented through cumulative parameters attached to nested time shells, while time-ordered connected observables collapse onto the smallest time in the correlator. A neural analogue is a causal long-horizon architecture or training objective whose multi-horizon interactions are parameterized by cumulative shell summaries rather than all pairwise horizon interactions. The mechanism is falsifiable through a collapse test: normalized connected cross-horizon correlations should become functions of the minimum horizon in approximately ballistic or weak-memory sequence dynamics.

Ideas from this paper

✓✓ Beats tuned baseline 2026

Time-Shell Long-Horizon Decoder

Replace dense pairwise interactions between all forecast horizons with nested time-shell summaries. For sorted horizons, the readout at shell j receives a cumulative embedding of all coefficients or queries assigned to later horizons, reproducing the paper's dependence on products such as \(\Pi_j=\prod_{l>j}e^{\alpha_l}=e^{\sum_{l>j}\alpha_l}\). This gives an \(O(Kd)\) multi-horizon interaction instead of an \(O(K^2d)\) temporal attention block and should work best for weak-memory…

Useful7/10
Difficulty5/10
Novelty8/10
Paper: Dynamical correlation functions of extensive charges after global quantum quenches arXiv:2607.19208