Test-time augmentation -- the aggregation of predictions across transformed
versions of a test input -- is a common practice in image classification.
Traditionally, predictions are combined using a simple average. In this paper,
we present 1) experimental analyses that shed light on cases in which the
simple average is suboptimal and 2) a method to address these shortcomings. A
key finding is that even when test-time augmentation produces a net improvement
in accuracy, it can change many correct predictions into incorrect predictions.
We delve into when and why test-time augmentation changes a prediction from
being correct to incorrect and vice versa. Building on these insights, we
present a learning-based method for aggregating test-time augmentations.
Experiments across a diverse set of models, datasets, and augmentations show
that our method delivers consistent improvements over existing approaches.

本论文提出了一种基于学习的测试时间增强方法，通过实验分析发现了传统简单平均的不足，在进行测试时间增强时，即使使准确率有所提高，也会使许多正确的预测变为错误的预测，探讨了这种现象的原因，并在各种模型、数据集和增强方法上验证了本方法的稳定性和准确性。

测试时间增强中更好的聚合

Better Aggregation in Test-Time Augmentation

The precise measurement of neutrino properties is among the highest
priorities in fundamental particle physics, involving many experiments
worldwide. Since the experiments rely on the interactions of neutrinos with
bound nucleons inside atomic nuclei, the planned advances in the scope and
precision of these experiments requires a commensurate effort in the
understanding and modeling of the hadronic and nuclear physics of these
interactions, which is incorporated as a nuclear model in neutrino event
generators. This model is essential to every phase of experimental analyses and
its theoretical uncertainties play an important role in interpreting every
result.
In this White Paper we discuss in detail the impact of neutrino-nucleus
interactions, especially the nuclear effects, on the measurement of neutrino
properties using the determination of oscillation parameters as a central
example. After an Executive Summary and a concise Overview of the issues, we
explain how the neutrino event generators work, what can be learned from
electron-nucleus interactions and how each underlying physics process - from
quasi-elastic to deep inelastic scattering - is understood today. We then
emphasize how our understanding must improve to meet the demands of future
experiments. With every topic we find that the challenges can be met only with
the active support and collaboration among specialists in strong interactions
and electroweak physics that include theorists and experimentalists from both
the nuclear and high energy physics communities.

本综述详细讨论了中微子与核相互作用，特别是核效应，对通过确定振荡参数来测量中微子性质的影响，并指出了各种潜在的物理过程及需要改进的方面。

中微子 - 核散射的现状和挑战：NuSTEC 白皮书

NuSTEC White Paper: Status and Challenges of Neutrino-Nucleus Scattering

We perform an extensive survey of non-standard Higgs decays that are
consistent with the 125 GeV Higgs-like resonance. Our aim is to motivate a
large set of new experimental analyses on the existing and forthcoming data
from the Large Hadron Collider (LHC). The explicit search for exotic Higgs
decays presents a largely untapped discovery opportunity for the LHC
collaborations, as such decays may be easily missed by other searches. We
emphasize that the Higgs is uniquely sensitive to the potential existence of
new weakly coupled particles and provide a unified discussion of a large class
of both simplified and complete models that give rise to characteristic
patterns of exotic Higgs decays. We assess the status of exotic Higgs decays
after LHC Run 1. In many cases we are able to set new nontrivial constraints by
reinterpreting existing experimental analyses. We point out that improvements
are possible with dedicated analyses and perform some preliminary collider
studies. We prioritize the analyses according to their theoretical motivation
and their experimental feasibility. This document is accompanied by a website
that will be continuously updated with further information:
this http URL

本文对非标准希格斯衰变进行了广泛调查，并着眼于现有和即将到来的大型强子对撞机数据上的新实验分析，重点搜索异质希格斯衰变，提供了一种非常好的发现机会，同时还评估了 LHC 运行 1 后异质希格斯衰变的状态，设置了新的非平凡约束条件，并重申了需要专门分析来改善实验条件。