Abstractive summarization has enjoyed renewed interest in recent years, thanks to pre-trained language models and the availability of large-scale datasets. Despite promising results, current models still suffer from generating factually inconsistent summaries, reducing their utility for real-world application. Several recent efforts attempt to address this by devising models that automatically detect factual inconsistencies in machine generated summaries. However, they focus exclusively on English, a language with abundant resources. In this work, we leverage factual consistency evaluation models to improve multilingual summarization. We explore two intuitive approaches to mitigate hallucinations based on the signal provided by a multilingual NLI model, namely data filtering and controlled generation. Experimental results in the 45 languages from the XLSum dataset show gains over strong baselines in both automatic and human evaluation.

Many real-world applications of language models (LMs), such as writing assistance and code autocomplete, involve human-LM interaction. However, most benchmarks are non-interactive in that a model produces output without human involvement. To evaluate human-LM interaction, we develop a new framework, Human-AI Language-based Interaction Evaluation (HALIE), that defines the components of interactive systems and dimensions to consider when designing evaluation metrics. Compared to standard, non-interactive evaluation, HALIE captures (i) the interactive process, not only the final output; (ii) the first-person subjective experience, not just a third-party assessment; and (iii) notions of preference beyond quality (e.g., enjoyment and ownership). We then design five tasks to cover different forms of interaction: social dialogue, question answering, crossword puzzles, summarization, and metaphor generation. With four state-of-the-art LMs (three variants of OpenAI's GPT-3 and AI21 Labs' Jurassic-1), we find that better non-interactive performance does not always translate to better human-LM interaction. In particular, we highlight three cases where the results from non-interactive and interactive metrics diverge and underscore the importance of human-LM interaction for LM evaluation.

In this paper, we investigate on improving the adversarial robustness obtained in adversarial training (AT) via reducing the difficulty of optimization. To better study this problem, we build a novel Bregman divergence perspective for AT, in which AT can be viewed as the sliding process of the training data points on the negative entropy curve. Based on this perspective, we analyze the learning objectives of two typical AT methods, i.e., PGD-AT and TRADES, and we find that the optimization process of TRADES is easier than PGD-AT for that TRADES separates PGD-AT. In addition, we discuss the function of entropy in TRADES, and we find that models with high entropy can be better robustness learners. Inspired by the above findings, we propose two methods, i.e., FAIT and MER, which can both not only reduce the difficulty of optimization under the 10-step PGD adversaries, but also provide better robustness. Our work suggests that reducing the difficulty of optimization under the 10-step PGD adversaries is a promising approach for enhancing the adversarial robustness in AT.

Fox News Flash top headlines are here. Check out what's clicking on Foxnews.com. The Islamic State of Iraq and Syria (ISIS) is claiming responsibility for the suicide bomb attacks in Iran this week, Fox News Digital has learned. ISIS claims to have orchestrated the double suicide bomber attack at the memorial to deceased Iranian military official Qassem Soleimani. A statement from ISIS published to Telegram named terrorist operatives Omar al-Mowahid and Sayefulla al-Mujahid as the suicidal attackers behind the "dual martyrdom operation."

The explosions that killed more than 100 people in Iran on Wednesday took place at an anniversary commemoration for Maj. Gen. Qassim Suleimani, the top Iranian commander who was killed by a U.S. drone strike four years ago. General Suleimani, the most powerful Iranian commander at the head of the foreign-facing arm of Iran's Revolutionary Guards Corps, was considered a hero by some in Iran and in other parts of the region for building an axis of allied militias to defend Iran's interests across the Middle East, to counter the United States and Israel, and for helping to defeat the Islamic State in Syria and Iraq. In the United States, he was regarded as a force behind international terrorism campaigns, and President Donald Trump said his killing in January 2020 was ordered "to stop a war" because General Suleimani had been plotting attacks on American diplomats and military personnel. General Suleimani was designated as a terrorist by the United States and Israel, where he helped orchestrate waves of militia attacks.

Retrieval-based augmentations that aim to incorporate knowledge from an external database into language models have achieved great success in various knowledge-intensive (KI) tasks, such as question-answering and text generation. However, integrating retrievals in non-knowledge-intensive (NKI) tasks, such as text classification, is still challenging. Existing works focus on concatenating retrievals to inputs as context to form the prompt-based inputs. Unfortunately, such methods require language models to have the capability to handle long texts. Besides, inferring such concatenated data would also consume a significant amount of computational resources. To solve these challenges, we propose \textbf{ReFusion} in this paper, a computation-efficient \textbf{Re}trieval representation \textbf{Fusion} with neural architecture search. The main idea is to directly fuse the retrieval representations into the language models. Specifically, we first propose an online retrieval module that retrieves representations of similar sentences. Then, we present a retrieval fusion module including two effective ranking schemes, i.e., reranker-based scheme and ordered-mask-based scheme, to fuse the retrieval representations with hidden states. Furthermore, we use Neural Architecture Search (NAS) to seek the optimal fusion structure across different layers. Finally, we conduct comprehensive experiments, and the results demonstrate our ReFusion can achieve superior and robust performance on various NKI tasks.

Crystalline materials, such as metals and semiconductors, nearly always contain a special defect type called dislocation. This defect decisively determines many important material properties, e.g., strength, fracture toughness, or ductility. Over the past years, significant effort has been put into understanding dislocation behavior across different length scales via experimental characterization techniques and simulations. This paper introduces the dislocation ontology (DISO), which defines the concepts and relationships related to linear defects in crystalline materials. We developed DISO using a top-down approach in which we start defining the most general concepts in the dislocation domain and subsequent specialization of them. DISO is published through a persistent URL following W3C best practices for publishing Linked Data. Two potential use cases for DISO are presented to illustrate its usefulness in the dislocation dynamics domain. The evaluation of the ontology is performed in two directions, evaluating the success of the ontology in modeling a real-world domain and the richness of the ontology.

Optical lithography is the main enabler to semiconductor manufacturing. It requires extensive processing to perform the Resolution Enhancement Techniques (RETs) required to transfer the design data to a working Integrated Circuits (ICs). The processing power and computational runtime for RETs tasks is ever increasing due to the continuous reduction of the feature size and the expansion of the chip area. State-of-the-art research sought Machine Learning (ML) technologies to reduce runtime and computational power, however they are still not used in production yet. In this study, we analyze the reasons holding back ML computational lithography from being production ready and present a novel highly scalable end-to-end flow that enables production ready ML-RET correction.

Foundational models with billions of parameters which have been trained on large corpora of data have demonstrated non-trivial skills in a variety of domains. However, due to their monolithic structure, it is challenging and expensive to augment them or impart new skills. On the other hand, due to their adaptation abilities, several new instances of these models are being trained towards new domains and tasks. In this work, we study the problem of efficient and practical composition of existing foundation models with more specific models to enable newer capabilities. To this end, we propose CALM -- Composition to Augment Language Models -- which introduces cross-attention between models to compose their representations and enable new capabilities. Salient features of CALM are: (i) Scales up LLMs on new tasks by 're-using' existing LLMs along with a few additional parameters and data, (ii) Existing model weights are kept intact, and hence preserves existing capabilities, and (iii) Applies to diverse domains and settings. We illustrate that augmenting PaLM2-S with a smaller model trained on low-resource languages results in an absolute improvement of up to 13\% on tasks like translation into English and arithmetic reasoning for low-resource languages. Similarly, when PaLM2-S is augmented with a code-specific model, we see a relative improvement of 40\% over the base model for code generation and explanation tasks -- on-par with fully fine-tuned counterparts.

Numerous developments in the recent literature show the promising potential of deep learning in obtaining numerical solutions to partial differential equations (PDEs) beyond the reach of current numerical solvers. However, data-driven neural operators all suffer from the same problem: the data needed to train a network depends on classical numerical solvers such as finite difference or finite element, among others. In this paper, we propose a new approach to generating synthetic functional training data that does not require solving a PDE numerically. The way we do this is simple: we draw a large number $N$ of independent and identically distributed `random functions' $u_j$ from the underlying solution space (e.g., $H_0^1(\Omega)$) in which we know the solution lies according to classical theory. We then plug each such random candidate solution into the equation and get a corresponding right-hand side function $f_j$ for the equation, and consider $(f_j, u_j)_{j=1}^N$ as supervised training data for learning the underlying inverse problem $f \rightarrow u$. This `backwards' approach to generating training data only requires derivative computations, in contrast to standard `forward' approaches, which require a numerical PDE solver, enabling us to generate a large number of such data points quickly and efficiently. While the idea is simple, we hope that this method will expand the potential for developing neural PDE solvers that do not depend on classical numerical solvers.