DeBERTa: Decoding-enhanced BERT with Disentangled Attention

Recent progress in pre-trained neural language models has significantly improved the performance of many natural language processing (NLP) tasks. In this paper we propose a new model architecture DeBERTa (Decoding-enhanced BERT with disentangled attention) that improves the BERT and RoBERTa models using two novel techniques. The first is the disentangled attention mechanism, where each word is represented using two vectors that encode its content and position, respectively, and the attention weights among words are computed using disentangled matrices on their contents and relative positions. Second, an enhanced mask decoder is used to replace the output softmax layer to predict the masked tokens for model pretraining. We show that these two techniques significantly improve the efficiency of model pre-training and performance of downstream tasks. Compared to RoBERTa-Large, a DeBERTa model trained on half of the training data performs consistently better on a wide range of NLP tasks, achieving improvements on MNLI by +0.9% (90.2% vs. 91.1%), on SQuAD v2.0 by +2.3% (88.4% vs. 90.7%) and RACE by +3.6% (83.2% vs. 86.8%). The DeBERTa code and pre-trained models will be made publicly available at https://github.com/microsoft/DeBERTa.

Adversarial Training for Commonsense Inference

We propose an AdversariaL training algorithm for commonsense InferenCE (ALICE). We apply small perturbations to word embeddings and minimize the resultant adversarial risk to regularize the model. We exploit a novel combination of two different approaches to estimate these perturbations: 1) using the true label and 2) using the model prediction. Without relying on any human-crafted features, knowledge bases, or additional datasets other than the target datasets, our model boosts the fine-tuning performance of RoBERTa, achieving competitive results on multiple reading comprehension datasets that require commonsense inference.

Adversarial Training for Large Neural Language Models

Generalization and robustness are both key desiderata for designing machine learning methods. Adversarial training can enhance robustness, but past work often finds it hurts generalization. In natural language processing (NLP), pre-training large neural language models such as BERT have demonstrated impressive gain in generalization for a variety of tasks, with further improvement from adversarial fine-tuning. However, these models are still vulnerable to adversarial attacks. In this paper, we show that adversarial pre-training can improve both generalization and robustness. We propose a general algorithm ALUM (Adversarial training for large neural LangUage Models), which regularizes the training objective by applying perturbations in the embedding space that maximizes the adversarial loss. We present the first comprehensive study of adversarial training in all stages, including pre-training from scratch, continual pre-training on a well-trained model, and task-specific fine-tuning. ALUM obtains substantial gains over BERT on a wide range of NLP tasks, in both regular and adversarial scenarios. Even for models that have been well trained on extremely large text corpora, such as RoBERTa, ALUM can still produce significant gains from continual pre-training, whereas conventional non-adversarial methods can not. ALUM can be further combined with task-specific fine-tuning to attain additional gains. The ALUM code is publicly available at https://github.com/namisan/mt-dnn.

Understanding the Difficulty of Training Transformers

Transformers have been proved effective for many deep learning tasks. Training transformers, however, requires non-trivial efforts regarding carefully designing learning rate schedulers and cutting-edge optimizers (the standard SGD fails to train Transformers effectively). In this paper, we study Transformer training from both theoretical and empirical perspectives. Our analysis reveals that unbalanced gradients are not the root cause of the instability of training. Instead, we identify an amplification effect that substantially influences training. Specifically, we observe that for each layer in a multi-layer Transformer model, heavy dependency on its residual branch makes training unstable since it amplifies small parameter perturbations (e.g., parameter updates) and result in significant disturbances in the model output, yet a light dependency limits the potential of model training and can lead to an inferior trained model. Inspired by our analysis, we propose Admin ($\mathbf{Ad}$aptive $\mathbf{m}$odel $\mathbf{in}$itialization) to stabilize the training in the early stage and unleash its full potential in the late stage. Extensive experiments show that Admin is more stable, converges faster, and leads to better performance.

UniLMv2: Pseudo-Masked Language Models for Unified Language Model Pre-Training

We propose to pre-train a unified language model for both autoencoding and partially autoregressive language modeling tasks using a novel training procedure, referred to as a pseudo-masked language model (PMLM). Given an input text with masked tokens, we rely on conventional masks to learn inter-relations between corrupted tokens and context via autoencoding, and pseudo masks to learn intra-relations between masked spans via partially autoregressive modeling. With well-designed position embeddings and self-attention masks, the context encodings are reused to avoid redundant computation. Moreover, conventional masks used for autoencoding provide global masking information, so that all the position embeddings are accessible in partially autoregressive language modeling. In addition, the two tasks pre-train a unified language model as a bidirectional encoder and a sequence-to-sequence decoder, respectively. Our experiments show that the unified language models pre-trained using PMLM achieve new state-of-the-art results on a wide range of natural language understanding and generation tasks across several widely used benchmarks.

The Microsoft Toolkit of Multi-Task Deep Neural Networks for Natural Language Understanding

We present MT-DNN, an open-source natural language understanding (NLU) toolkit that makes it easy for researchers and developers to train customized deep learning models. Built upon PyTorch and Transformers, MT-DNN is designed to facilitate rapid customization for a broad spectrum of NLU tasks, using a variety of objectives (classification, regression, structured prediction) and text encoders (e.g., RNNs, BERT, RoBERTa, UniLM). A unique feature of MT-DNN is its built-in support for robust and transferable learning using the adversarial multi-task learning paradigm. To enable efficient production deployment, MT-DNN supports multi-task knowledge distillation, which can substantially compress a deep neural model without significant performance drop. We demonstrate the effectiveness of MT-DNN on a wide range of NLU applications across general and biomedical domains. The software and pre-trained models will be publicly available at https://github.com/namisan/mt-dnn.

MLFcGAN: Multi-level Feature Fusion based Conditional GAN for Underwater Image Color Correction

Color correction for underwater images has received increasing interests, due to its critical role in facilitating available mature vision algorithms for underwater scenarios. Inspired by the stunning success of deep convolutional neural networks (DCNNs) techniques in many vision tasks, especially the strength in extracting features in multiple scales, we propose a deep multi-scale feature fusion net based on the conditional generative adversarial network (GAN) for underwater image color correction. In our network, multi-scale features are extracted first, followed by augmenting local features on each scale with global features. This design was verified to facilitate more effective and faster network learning, resulting in better performance in both color correction and detail preservation. We conducted extensive experiments and compared with the state-of-the-art approaches quantitatively and qualitatively, showing that our method achieves significant improvements.

Multi-Modality Cascaded Fusion Technology for Autonomous Driving

Multi-modality fusion is the guarantee of the stability of autonomous driving systems. In this paper, we propose a general multi-modality cascaded fusion framework, exploiting the advantages of decision-level and feature-level fusion, utilizing target position, size, velocity, appearance and confidence to achieve accurate fusion results. In the fusion process, dynamic coordinate alignment(DCA) is conducted to reduce the error between sensors from different modalities. In addition, the calculation of affinity matrix is the core module of sensor fusion, we propose an affinity loss that improves the performance of deep affinity network(DAN). Last, the proposed step-by-step cascaded fusion framework is more interpretable and flexible compared to the end-toend fusion methods. Extensive experiments on Nuscenes [2] dataset show that our approach achieves the state-of-theart performance.dataset show that our approach achieves the state-of-the-art performance.

RAT-SQL: Relation-Aware Schema Encoding and Linking for Text-to-SQL Parsers

When translating natural language questions into SQL queries to answer questions from a database, contemporary semantic parsing models struggle to generalize to unseen database schemas. The generalization challenge lies in (a) encoding the database relations in an accessible way for the semantic parser, and (b) modeling alignment between database columns and their mentions in a given query. We present a unified framework, based on the relation-aware self-attention mechanism, to address schema encoding, schema linking, and feature representation within a text-to-SQL encoder. On the challenging Spider dataset this framework boosts the exact match accuracy to 53.7%, compared to 47.4% for the state-of-the-art model unaugmented with BERT embeddings. In addition, we observe qualitative improvements in the model's understanding of schema linking and alignment.