ProtocolMatch: Protocol-Dependent Model Selection for Scientific Dynamics Forecasting
Abstract
Scientific dynamics forecasting is often framed as an architecture choice, although deployment is also determined by observed history, rollout feedback, compute budget, physical objective, and test distribution. We formulate protocol-dependent model selection and introduce ProtocolMatch, a compute-matched, validation-selected, and failure-preserving evaluation framework. On driven quantum-spin dynamics, we compare recurrent, patched-attention, causal-attention, and low-rank linear predictors across three independently generated datasets. The causal-attention--recurrence ordering reverses as the training set grows within a fixed two-spin task, while a linear predictor has the lowest mean error in the six-spin local-observable comparison. Restricting observed history worsens every refreshed-history view but improves every closed-loop view in the four-spin study. A latest-state MLP has lower error than persistence on every dataset under state refresh across all five cells, yet its closed-loop rank varies by system and includes finite explosive errors. Physical penalties improve targeted consistency without reliably improving prediction error, and in-distribution intervals lose most coverage after a driving-frequency shift. Thus scientific model selection should return a predictor with its protocol and report accuracy, physical validity, and shifted-distribution reliability separately.
Related
- Hardware-Aware Design and Layout Selection in Quantum Reservoir Computing for Insurance Claims Forecasting
- A self-tallying quantum anonymous voting protocol for multiple-selection elections
- A Centre-Charge Selection Rule for the Wilson-Line Potential: The Fundamental Domain of Gauge-Higgs Unification Is Representation-Dependent (Part III)
- Landscape-Dependent Performance of Photonic Quantum Solvers in QUBO Feature Selection for Financial Risk Detection
- Dimension-dependent order parameter selection in higher-order Kuramoto dynamics on spheres: continuous versus quantized regimes
BibTeX
@misc{wei2026protocolmatch,
title = {{ProtocolMatch: Protocol-Dependent Model Selection for Scientific Dynamics Forecasting}},
author = {Lu Wei and Yufeng Wang and Haibin Ling},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.10239},
eprint = {2610.10239},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.10239},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - ProtocolMatch: Protocol-Dependent Model Selection for Scientific Dynamics Forecasting
AU - Lu Wei
AU - Yufeng Wang
AU - Haibin Ling
PY - 2026
DA - 2026/10/07
PB - arXiv
DO - 10.48550/arxiv.2610.10239
UR - https://arxiv.org/abs/2610.10239
AB - Scientific dynamics forecasting is often framed as an architecture choice, although deployment is also determined by observed history, rollout feedback, compute budget, physical objective, and test distribution. We formulate protocol-dependent model selection and introduce ProtocolMatch, a compute-matched, validation-selected, and failure-preserving evaluation framework. On driven quantum-spin dynamics, we compare recurrent, patched-attention, causal-attention, and low-rank linear predictors across three independently generated datasets. The causal-attention--recurrence ordering reverses as the training set grows within a fixed two-spin task, while a linear predictor has the lowest mean error in the six-spin local-observable comparison. Restricting observed history worsens every refreshed-history view but improves every closed-loop view in the four-spin study. A latest-state MLP has lower error than persistence on every dataset under state refresh across all five cells, yet its closed-loop rank varies by system and includes finite explosive errors. Physical penalties improve targeted consistency without reliably improving prediction error, and in-distribution intervals lose most coverage after a driving-frequency shift. Thus scientific model selection should return a predictor with its protocol and report accuracy, physical validity, and shifted-distribution reliability separately.
ER -