Controllable text generation is an appealing but challenging task, which allows users to specify particular attributes of the generated outputs. In this paper, we propose a controllable dialogue generation model to steer response generation under mul
ti-attribute constraints. Specifically, we define and categorize the commonly used control attributes into global and local ones, which possess different granularities of effects on response generation. Then, we significantly extend the conventional seq2seq framework by introducing a novel two-stage decoder, which first uses a multi-grained style specification layer to impose the stylistic constraints and determine word-level control states of responses based on the attributes, and then employs a response generation layer to generate final responses maintaining both semantic relevancy to the contexts and fidelity to the attributes. Furthermore, we train our model with an attribute consistency reward to promote response control with explicit supervision signals. Extensive experiments and in-depth analyses on two datasets indicate that our model can significantly outperform competitive baselines in terms of response quality, content diversity and controllability.
Within an isospin- and momentum-dependent transport model, we investigate the necessity of selfconsistent calculations for the electromagnetic field in probing the nuclear symmetry energy using pion observables in heavy-ion collisions at intermediate
energies. To this end, we perform the $^{96}$Ru + $^{96}$Ru collisions at 400 MeV/nucleon with two calculations scenarios for the electromagnetic field including the selfconsistent calculation and the most used Li{e}nard-Wiechert formula, while the latter is a simplified one of the complete Li{e}nard-Wiechert formula by neglecting the radiation field for practical calculations in heavy-ion collisions at intermediate and/or relativistic energies. As a comparison, we also consider the static Coulomb field formula for calculations of the electromagnetic field in heavy-ion collisions. It is shown that the most used simplified Li{e}nard-Wiechert formula is not enough for the electromagnetic field calculation because the absent radiation field in this formula also affects significantly the charged pions as well as their $pi^{-}/pi^{+}$ ratio. Moreover, we also examine effects of the electromagnetic field in these scenarios on the double $pi^{-}/pi^{+}$ ratio of two isobar reaction systems of $^{96}$Ru + $^{96}$Ru and $^{96}$Zr + $^{96}$Zr at 400 MeV/nucleon. It is shown that the double $pi^{-}/pi^{+}$ ratio of two reactions tends to be less affected by the electromagnetic field calculation scenario and thus can still be an effective probe of the nuclear symmetry energy in heavy-ion collisions. Therefore, according to these findings, it is suggested that the selfconsistent calculation for the electromagnetic field should be carefully taken into account when using the pion observables to probe the nuclear symmetry energy in heavy-ion collisions.
Graph neural networks (GNNs) have been widely used in deep learning on graphs. They can learn effective node representations that achieve superior performances in graph analysis tasks such as node classification and node clustering. However, most met
hods ignore the heterogeneity in real-world graphs. Methods designed for heterogeneous graphs, on the other hand, fail to learn complex semantic representations because they only use meta-paths instead of meta-graphs. Furthermore, they cannot fully capture the content-based correlations between nodes, as they either do not use the self-attention mechanism or only use it to consider the immediate neighbors of each node, ignoring the higher-order neighbors. We propose a novel Higher-order Attribute-Enhancing (HAE) framework that enhances node embedding in a layer-by-layer manner. Under the HAE framework, we propose a Higher-order Attribute-Enhancing Graph Neural Network (HAEGNN) for heterogeneous network representation learning. HAEGNN simultaneously incorporates meta-paths and meta-graphs for rich, heterogeneous semantics, and leverages the self-attention mechanism to explore content-based nodes interactions. The unique higher-order architecture of HAEGNN allows examining the first-order as well as higher-order neighborhoods. Moreover, HAEGNN shows good explainability as it learns the importances of different meta-paths and meta-graphs. HAEGNN is also memory-efficient, for it avoids per meta-path based matrix calculation. Experimental results not only show HAEGNN superior performance against the state-of-the-art methods in node classification, node clustering, and visualization, but also demonstrate its superiorities in terms of memory efficiency and explainability.
In late-2020, many countries around the world faced another surge in number of confirmed cases of COVID-19, including United Kingdom, Canada, Brazil, United States, etc., which resulted in a large nationwide and even worldwide wave. While there have
been indications that precaution fatigue could be a key factor, no scientific evidence has been provided so far. We used a stochastic metapopulation model with a hierarchical structure and fitted the model to the positive cases in the US from the start of outbreak to the end of 2020. We incorporated non-pharmaceutical interventions (NPIs) into this model by assuming that the precaution strength grows with positive cases and studied two types of pandemic fatigue. We found that people in most states and in the whole US respond to the outbreak in a sublinear manner (with exponent k=0.5), while only three states (Massachusetts, New York and New Jersey) have linear reaction (k=1). Case fatigue (decline in peoples vigilance to positive cases) is responsible for 58% of cases, while precaution fatigue (decay of maximal fraction of vigilant group) accounts for 26% cases. If there were no pandemic fatigue (no case fatigue and no precaution fatigue), total positive cases would have reduced by 68% on average. Our study shows that pandemic fatigue is the major cause of the worsening situation of COVID-19 in United States. Reduced vigilance is responsible for most positive cases, and higher mortality rate tends to push local people to react to the outbreak faster and maintain vigilant for longer time.
Unsupervised cross-domain person re-identification (Re-ID) aims to adapt the information from the labelled source domain to an unlabelled target domain. Due to the lack of supervision in the target domain, it is crucial to identify the underlying sim
ilarity-and-dissimilarity relationships among the unlabelled samples in the target domain. In order to use the whole data relationships efficiently in mini-batch training, we apply a series of memory modules to maintain an up-to-date representation of the entire dataset. Unlike the simple exemplar memory in previous works, we propose a novel multi-level memory network (MMN) to discover multi-level complementary information in the target domain, relying on three memory modules, i.e., part-level memory, instance-level memory, and domain-level memory. The proposed memory modules store multi-level representations of the target domain, which capture both the fine-grained differences between images and the global structure for the holistic target domain. The three memory modules complement each other and systematically integrate multi-level supervision from bottom to up. Experiments on three datasets demonstrate that the multi-level memory modules cooperatively boost the unsupervised cross-domain Re-ID task, and the proposed MMN achieves competitive results.
We experimentally identify coherent spin pumping in the magnon-magnon hybrid modes of permalloy/yttrium iron garnet (Py/YIG) bilayers. Using broadband ferromagnetic resonance, an avoided crossing is observed between the uniform mode of Py and the spi
n wave mode of YIG due to the fieldlike interfacial exchange coupling. We also identify additional linewidth suppression and enhancement for the in-phase and out-of-phase hybrid modes, respectively, textcolor{black}{which can be interpreted as concerted dampinglike torque from spin pumping}. Our analysis predicts inverse proportionality of both fieldlike and dampinglike torques to the square root of the Py thickness, which quantitatively agrees with experiments.
Vehicle instance retrieval often requires one to recognize the fine-grained visual differences between vehicles. Besides the holistic appearance of vehicles which is easily affected by the viewpoint variation and distortion, vehicle parts also provid
e crucial cues to differentiate near-identical vehicles. Motivated by these observations, we introduce a Part-Guided Attention Network (PGAN) to pinpoint the prominent part regions and effectively combine the global and part information for discriminative feature learning. PGAN first detects the locations of different part components and salient regions regardless of the vehicle identity, which serve as the bottom-up attention to narrow down the possible searching regions. To estimate the importance of detected parts, we propose a Part Attention Module (PAM) to adaptively locate the most discriminative regions with high-attention weights and suppress the distraction of irrelevant parts with relatively low weights. The PAM is guided by the instance retrieval loss and therefore provides top-down attention that enables attention to be calculated at the level of car parts and other salient regions. Finally, we aggregate the global appearance and part features to improve the feature performance further. The PGAN combines part-guided bottom-up and top-down attention, global and part visual features in an end-to-end framework. Extensive experiments demonstrate that the proposed method achieves new state-of-the-art vehicle instance retrieval performance on four large-scale benchmark datasets.
Laser-induced electron tunneling underlies numerous emerging spectroscopic techniques to probe attosecond electron dynamics in atoms and molecules. The improvement of those techniques requires an accurate knowledge of the exit momentum for the tunnel
ing wave packet. Here we demonstrate a photoelectron interferometric scheme to probe the electron momentum longitudinal to the tunnel direction at the tunnel exit by measuring the photoelectron holographic pattern in an orthogonally polarized two-color laser pulse. In this scheme, we use a perturbative 400-nm laser field to modulate the photoelectron holographic fringes generated by a strong 800-nm pulse. The fringe shift offers a direct experimental access to the intermediate canonical momentum of the rescattering electron, allowing us to reconstruct the momentum offset at the tunnel exit with high accuracy. Our result unambiguously proves the existence of nonzero initial longitudinal momentum at the tunnel exit and provides fundamental insights into the non-quasi-static nature of the strong-field tunneling.
Generally, lattice distortions play a key role in determining the ground states of materials. Although it is well known that trigonal distortions are generic to most two-dimensional transition metal dichalcogenides, the impact of this structural dist
ortion on the electronic structure has not been understood conclusively. Here, by using a combination of polarization dependent X-ray absorption spectroscopy (XAS), X-ray photoelectron spectroscopy (XPS) and atomic multiplet cluster calculations, we have investigated the electronic structure of titanium dichalcogenides TiX2 (X=S, Se, Te), where the magnitude of the trigonal distortion increase monotonically from S to Se and Te. Our results reveal the presence of an anomalous and large crystal filed splitting. This unusual kind of crystal field splitting is likely responsible for the unconventional electronic structure of TiX2 compounds. Our results also indicate the drawback of the distorted crystal field picture in explaining the observed electronic ground state of these materials and emphasize the key importance of metal-ligand hybridization and electronic correlation in defining the electronic structures near Fermi energy.
Polar metals, commonly defined by the coexistence of polar crystal structure and metallicity, are thought to be scarce because the long-range electrostatic fields favoring the polar structure are expected to be fully screened by the conduction electr
ons of a metal. Moreover, reducing from three to two dimensions, it remains an open question whether a polar metal can exist. Here we report on the realization of a room temperature two-dimensional polar metal of the B-site type in tri-color (tri-layer) superlattices BaTiO$_3$/SrTiO$_3$/LaTiO$_3$. A combination of atomic resolution scanning transmission electron microscopy with electron energy loss spectroscopy, optical second harmonic generation, electrical transport, and first-principles calculations have revealed the microscopic mechanisms of periodic electric polarization, charge distribution, and orbital symmetry. Our results provide a route to creating all-oxide artificial non-centrosymmetric quasi-two-dimensional metals with exotic quantum states including coexisting ferroelectric, ferromagnetic, and superconducting phases.
