Question answering (QA) models often rely on large-scale training datasets,
which necessitates the development of a data generation framework to reduce the
cost of manual annotations. Although several recent studies have aimed to
generate synthetic questions with single-span answers, no study has been
conducted on the creation of list questions with multiple, non-contiguous spans
as answers. To address this gap, we propose LIQUID, an automated framework for
generating list QA datasets from unlabeled corpora. We first convert a passage
from Wikipedia or PubMed into a summary and extract named entities from the
summarized text as candidate answers. This allows us to select answers that are
semantically correlated in context and is, therefore, suitable for constructing
list questions. We then create questions using an off-the-shelf question
generator with the extracted entities and original passage. Finally, iterative
filtering and answer expansion are performed to ensure the accuracy and
completeness of the answers. Using our synthetic data, we significantly improve
the performance of the previous best list QA models by exact-match F1 scores of
5.0 on MultiSpanQA, 1.9 on Quoref, and 2.8 averaged across three BioASQ
benchmarks.

LIQUID 是一个自动化框架，用于从未标注的文本中生成列表问答数据集，并可显著提高之前在 MultiSpanQA、Quoref 和 BioASQ 数据集上表现得最好的列表问答模型的精确匹配 F1 分数。

LIQUID: 列表型问题回答数据集生成框架

LIQUID: A Framework for List Question Answering Dataset Generation

Mass-spring models are essential for the description of sloshing resonances
in engineering. By experimentally measuring the liquid's centre of mass in a
horizontally oscillated rectangular tank, we show that low-amplitude sloshing
obeys the Duffing equation. A bending of the response curve in analogy to a
softening spring is observed, with growing hysteresis as the driving amplitude
increases. At large amplitudes, complex wave patterns emerge (including
wave-breaking and run up at the tank walls), competition between flow states is
observed and the dynamics departs progressively from Duffing. We also provide a
quantitative comparison of wave shapes and response curves to the predictions
of a multimodal model based on potential flow theory (Faltinsen & Timokha 2009)
and show that it systematically overestimates the sloshing amplitudes and the
hysteresis. We find that the phase-lag between the liquid's centre of mass and
the forcing is the key predictor of the nonlinear response maxima. The
phase-lag reflects precisely the onset of deviations from Duffing dynamics and
- most importantly - at resonance the sloshing motion always lags the driving
by 90{\deg} (independently of the wave pattern). This confirms the theoretical
90{\deg}-phase-lag criterion (Cenedese & Haller 2020).

本研究实验测量了矩形水槽中液体的质心，并表明低幅度晃荡服从杜芬方程，且响应曲线弯曲类似于软化弹簧，随着驱动幅度的增加，它们具有增长的滞后，并且在幅度大时，会出现复杂的波动模式和与流态之间的竞争。最后，实验发现相位滞后是非线性响应 maxima 的关键预测变量，并且在共振时，晃动运动始终滞后于驱动 90 度。