Adversarial attacks are carried out to reveal the vulnerability of deep neural networks. Textual adversarial attacking is challenging because text is discrete and a small perturbation can bring significant change to the original input. Word-level attacking, which can be regarded as a combinatorial optimization problem, is a well-studied class of textual attack methods. However, existing word-level attack models are far from perfect, largely because unsuitable search space reduction methods and inefficient optimization algorithms are employed. In this paper, we propose a novel attack model, which incorporates the sememebased word substitution method and particle swarm optimization-based search algorithm to solve the two problems separately. We conduct exhaustive experiments to evaluate our attack model by attacking BiLSTM and BERT on three benchmark datasets. Experimental results demonstrate that our model consistently achieves much higher attack success rates and crafts more high-quality adversarial examples as compared to baseline methods. Also, further experiments show our model has higher transferability and can bring more robustness enhancement to victim models by adversarial training. All the code and data of this paper can be obtained on https://github.com/ thunlp/SememePSO-Attack. * Indicates equal contribution. Yuan developed the method, designed and conducted most experiments; Fanchao formalized the task, designed some experiments and wrote the paper; Chenghao made the original research proposal, performed human evaluation and conducted some experiments.
Myasthenia gravis (MG) is a chronic humoral immunity-mediated autoimmune disorder of the neuromuscular junction characterized by muscle weakness. Follicular helper T (Tfh) cells may be the key Th cell subset that promotes MG development, as their major function is helping B cell activation and Ab production. Aberrance of thymus-derived Tfh cells might be implicated in autoimmune diseases including MG; just how circulating Tfh cells, especially those from patients with a normal thymus, contribute to MG pathogenesis remains to be uncovered. In this article, we characterize a population of circulating CD4(+)CXCR5(+)PD-1(+) Tfh cells in ocular and generalized MG patients without thymic abnormalities and demonstrate that the circulating Tfh cells are significantly enriched in generalized MG patients but not in ocular MG patients compared with healthy subjects, whereas a proportion of follicular regulatory T cells decreased in MG patients. In addition, the frequency of plasma cells and B cells was higher and the serum levels of IL-6/IL-21 were also elevated in these MG patients. The activated Tfh1 and Tfh17 in Tfh cells are the major source for IL-21 production in MG patients. A strong correlation between Tfh cells and the plasma cell frequency and anti-acetylcholine receptor Ab titers was evident in generalized MG patients. In particular, we found that Tfh cells derived from MG patients promoted B cells to produce Abs in an IL-21 signaling-dependent manner. Collectively, our results suggest that circulating Tfh cells may act on autoreactive B cells and thus contribute to the development of MG in patients without thymic abnormalities.
As the water-gas shift (WGS) reaction serves as a crucial industrial process, strategies for developing robust WGS catalysts are highly desiderated. Here we report the construction of stabilized bulk-nano interfaces to fabricate highly efficient copper-ceria catalyst for the WGS reaction. With an in-situ structural transformation, small CeO 2 nanoparticles (2–3 nm) are stabilized on bulk Cu to form abundant CeO 2 -Cu interfaces, which maintain well-dispersed under reaction conditions. This inverse CeO 2 /Cu catalyst shows excellent WGS performances, of which the activity is 5 times higher than other reported Cu catalysts. Long-term stability is also very solid under harsh conditions. Mechanistic study illustrates that for the inverse CeO 2 /Cu catalyst, superb capability of H 2 O dissociation and CO oxidation facilitates WGS process via the combination of associative and redox mechanisms. This work paves a way to fabricate robust catalysts by combining the advantages of bulk and nano-sized catalysts. Catalysts with such inverse configurations show great potential in practical WGS applications.
Continuum robots exhibit promising adaptability and dexterity for soft manipulation due to their intrinsic compliance. However, this compliance may lead to challenges in modeling as well as positioning and loading. In this paper, a virtual work-based static model is established to describe the deformation and mechanics of continuum robots with a generic rod-driven structure, taking the geometric constraint of the drive rods into account. Following this, this paper presents a novel variable stiffness mechanism powered by a set of embedded Shape Memory Alloy (SMA) springs, which can make the drive rods become ‘locked’ on the body structure with different configurations. The resulting effects of variable stiffness are then presented in the static model by introducing tensions of the SMA and friction on the rods. Compared with conventional models, there is no need to predefine the actuation forces of the drive rods; instead, actuation displacements are used in this new mechanism system with stiffness being regulated. As a result, the phenomenon that the continuum robot can exhibit an S-shaped curve when subject to single-directional forces is observed and analyzed. Simulations and experiments demonstrated that the presented mechanism has stiffness variation of over 287% and further demonstrated that the mechanism and its model are achievable with good accuracy, such that the ratio of positioning error is less than 2.23% at the robot end-effector to the robot length.
Semantic compositionality (SC) refers to the phenomenon that the meaning of a complex linguistic unit can be composed of the meanings of its constituents. Most related works focus on using complicated compositionality functions to model SC while few works consider external knowledge in models. In this paper, we verify the effectiveness of sememes, the minimum semantic units of human languages, in modeling SC by a confirmatory experiment. Furthermore, we make the first attempt to incorporate sememe knowledge into SC models, and employ the sememeincorporated models in learning representations of multiword expressions, a typical task of SC. In experiments, we implement our models by incorporating knowledge from a famous sememe knowledge base HowNet and perform both intrinsic and extrinsic evaluations. Experimental results show that our models achieve significant performance boost as compared to the baseline methods without considering sememe knowledge. We further conduct quantitative analysis and case studies to demonstrate the effectiveness of applying sememe knowledge in modeling SC. All the code and data of this paper can be obtained on https: //github.com/thunlp/Sememe-SC.
To study the tumor specificity and antitumor activity of the replication-competent oncolytic adenovirus TOA02, which is controlled by a modified human telomerase reverse transcriptase (hTERT) promoter and expresses granulocyte macrophage colony-stimulating factor (GM-CSF). The wild-type hTERT promoter was modified, by inserting two E2F-binding sites. The effect of the modified hTERT on the viral yield and cytotoxicity of TOA02 were determined in vitro with a panel of tumor cells and normal cells, to evaluate tumor specificity; the effect on the antitumor efficacy and toxicity of TOA02 were determined in vivo, to evaluate the therapeutic potential of the adenovirus. The TOA02 adenovirus, which contained the modified hTERT promoter, produced a higher yield of virus in telomerase-positive and retinoblastoma-defective human cells, and a lower yield of virus in normal human cells than the wild-type adenovirus. A single injection of TOA02 showed strong antitumor efficacy in nude mice with human head/neck and hepatocellular carcinoma xenografts, and the efficacy further improved when used in combination with chemotherapy and with different routes of administration and regimens. In immunocompetent mice, the addition of GM-CSF produced a stronger antitumor activity and induced more mature dendritic cells and macrophages. The TOA02 adenovirus showed strong tumor-cell selectivity in vitro and antitumor efficacy in mouse models of human head/neck and hepatocellular cancer, suggesting that TOA02 has potential clinical applications for the treatment of solid tumors.
A lot of progress has been made to improve question answering (QA) in recent years, but the special problem of QA over narrative book stories has not been explored in-depth. We formulate BookQA as an open-domain QA task given its similar dependency on evidence retrieval. We further investigate how state-ofthe-art open-domain QA approaches can help BookQA. Besides achieving state-of-the-art on the NarrativeQA benchmark, our study also reveals the difficulty of evidence retrieval in books with a wealth of experiments and analysis -which necessitates future effort on novel solutions for evidence retrieval in BookQA.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.