Search

Summarizing lengthy documents is a common and essential task in our daily lives. Although recent advancements in neural summarization models can assist in crafting general-purpose summaries, human writers often have specific requirements that call for a more customized approach. To address this need, we introduce REVISE (Refinement and Editing via Iterative Summarization Enhancement), an innovative framework designed to facilitate iterative editing and refinement of draft summaries by human writers. Within our framework, writers can effortlessly modify unsatisfactory segments at any location or length and provide optional starting phrases -- our system will generate coherent alternatives that seamlessly integrate with the existing summary. At its core, REVISE incorporates a modified fill-in-the-middle model with the encoder-decoder architecture while developing novel evaluation metrics tailored for the summarization task. In essence, our framework empowers users to create high-quality, personalized summaries by effectively harnessing both human expertise and AI capabilities, ultimately transforming the summarization process into a truly collaborative and adaptive experience.

Large Language Models (LLMs) play a powerful \textit{Reader} of the \textit{Retrieve-then-Read} pipeline, making great progress in knowledge-based open-domain tasks. This work introduces a new framework, \textit{Rewrite-Retrieve-Read} that improves the retrieval-augmented method from the perspective of the query rewriting. Prior studies mostly contribute to adapt the retriever or stimulate the reader. Different from them, our approach pay attention of the query adaptation. Because the original query can not be always optimal to retrieve for the LLM, especially in the real world.(1) We first prompt an LLM to rewrite the queries, then conduct retrieval-augmented reading. (2) We further apply a small language model as a trainable rewriter, which rewrite the search query to cater to the frozen retriever and the LLM reader. To fine-tune the rewriter, we first use a pseudo data to conduct supervised warm-up training. Then the \textit{Retrieve-then-Read} pipeline is modeled as a reinforcement learning context. The rewriter is further trained as a policy model by maximize the reward of the pipeline performance. Evaluation is performed on two downstream tasks, open-domain QA and multiple choice. Our framework is proved effective and scalable., working in progress

Based on the remarkable achievements of pre-trained language models in abstractive summarization, the copying mechanism has proved helpful by improving the factuality, stability, and overall performance. This work proposes PROM, a new PhRase-level cOpying Mechanism that enhances attention on n-grams, which can be applied to zero-shot summarization with pre-training. PROM adds an indicator layer to explicitly pick up tokens in n-gram that can be copied from the source, and calculates an auxiliary loss for the copying prediction. Empirical studies show that PROM makes significant improvements in fine-tuning on benchmarks. In zero-shot setting, PROM is utilized in the self-supervised pre-training on raw corpora and provides new general baselines on a wide range of summarization datasets. Further analysis shows that PROM performs more reasonable copying and contributes to faithfulness.

Many applications of text generation such as summarization benefit from accurately controlling the text length. Existing approaches on length-controlled summarization either result in degraded performance or can only control the length approximately. In this work, we present a framework to generate summaries with precisely the specified number of tokens or sentences, while maintaining or even improving the text quality. In addition, we jointly train the models to predict the lengths, so our model can generate summaries with optimal length. We evaluate the proposed framework on the CNNDM dataset and show improved performance compared to existing methods.

Fine-tuning large pre-trained language models on downstream tasks has become an important paradigm in NLP. However, common practice fine-tunes all of the parameters in a pre-trained model, which becomes prohibitive when a large number of downstream tasks are present. Therefore, many fine-tuning methods are proposed to learn incremental updates of pre-trained weights in a parameter efficient way, e.g., low-rank increments. These methods often evenly distribute the budget of incremental updates across all pre-trained weight matrices, and overlook the varying importance of different weight parameters. As a consequence, the fine-tuning performance is suboptimal. To bridge this gap, we propose AdaLoRA, which adaptively allocates the parameter budget among weight matrices according to their importance score. In particular, AdaLoRA parameterizes the incremental updates in the form of singular value decomposition. Such a novel approach allows us to effectively prune the singular values of unimportant updates, which is essentially to reduce their parameter budget but circumvent intensive exact SVD computations. We conduct extensive experiments with several pre-trained models on natural language processing, question answering, and natural language generation to validate the effectiveness of AdaLoRA. Results demonstrate that AdaLoRA manifests notable improvement over baselines, especially in the low budget settings. Our code is publicly available at https://github.com/QingruZhang/AdaLoRA ., The 11th International Conference on Learning Representations (ICLR 2023)

Purpose: To examine the role of miR-557 in the pathogenesis and development of glioblastoma (GBM) and its potential regulatory mechanism. Methods: Dysregulation of miR-557 in GBM cells was determined using quantitative (real-time) PCR (qPCR). Cell proliferative ability, migratory activity and invasive ability were measured using the MTT colorimetric assay, wound healing assay and Transwell chambers assay. The direct target mRNA of miR-557 was predicted, and further validated using TargetScan and luciferase assay, respectively. Results: Compared with human astrocytes (HAs), the downregulation of miR-557 was observed in human GBM cell lines. In U-251MG and A172 cells that overexpressed miR-557 mimics, the level of proliferation was significantly reduced, the wound width was larger, and the number of invaded cells wan decreased than that of NC mimics. Predictive results from TargetScan and results of luciferase assay demonstrated that miR-557 directly targets the 3’-untranslated region (3’-UTR) of ADAM17. In A172 cells and U-251MG cells, the upregulation E-cadherin (E-cad) and the downregulation of N- cadherin (N-cad) and vimentin (Vim) were caused by miR-557 mimics as compared with NC mimics. Expression of ADAM17, NICD, HES1, and EGFR was downregulated by miR-557 mimics and upregulated by miR-557 inhibitors in A172 cells and U-251MG cells. Conclusion: Downregulation of miR-557 accelerates GBM cell growth and malignancy by directly targeting ADAM17 3’-UTR, providing a prognostic and potential therapeutic factor for new drug discovery in GBM.

Large language models (LLMs), such as ChatGPT, are able to generate human-like, fluent responses for many downstream tasks, e.g., task-oriented dialog and question answering. However, applying LLMs to real-world, mission-critical applications remains challenging mainly due to their tendency to generate hallucinations and their inability to use external knowledge. This paper proposes a LLM-Augmenter system, which augments a black-box LLM with a set of plug-and-play modules. Our system makes the LLM generate responses grounded in external knowledge, e.g., stored in task-specific databases. It also iteratively revises LLM prompts to improve model responses using feedback generated by utility functions, e.g., the factuality score of a LLM-generated response. The effectiveness of LLM-Augmenter is empirically validated on two types of scenarios, task-oriented dialog and open-domain question answering. LLM-Augmenter significantly reduces ChatGPT's hallucinations without sacrificing the fluency and informativeness of its responses. We make the source code and models publicly available., 15 pages

Unsupervised clustering under domain shift (UCDS) studies how to transfer the knowledge from abundant unlabeled data from multiple source domains to learn the representation of the unlabeled data in a target domain. In this paper, we introduce Prototype-oriented Clustering with Distillation (PCD) to not only improve the performance and applicability of existing methods for UCDS, but also address the concerns on protecting the privacy of both the data and model of the source domains. PCD first constructs a source clustering model by aligning the distributions of prototypes and data. It then distills the knowledge to the target model through cluster labels provided by the source model while simultaneously clustering the target data. Finally, it refines the target model on the target domain data without guidance from the source model. Experiments across multiple benchmarks show the effectiveness and generalizability of our source-private clustering method.

Transformer models have achieved remarkable results in various natural language tasks, but they are often prohibitively large, requiring massive memories and computational resources. To reduce the size and complexity of these models, we propose LoSparse (Low-Rank and Sparse approximation), a novel model compression technique that approximates a weight matrix by the sum of a low-rank matrix and a sparse matrix. Our method combines the advantages of both low-rank approximations and pruning, while avoiding their limitations. Low-rank approximation compresses the coherent and expressive parts in neurons, while pruning removes the incoherent and non-expressive parts in neurons. Pruning enhances the diversity of low-rank approximations, and low-rank approximation prevents pruning from losing too many expressive neurons. We evaluate our method on natural language understanding, question answering, and natural language generation tasks. We show that it significantly outperforms existing compression methods.

Diffusion models are powerful, but they require a lot of time and data to train. We propose Patch Diffusion, a generic patch-wise training framework, to significantly reduce the training time costs while improving data efficiency, which thus helps democratize diffusion model training to broader users. At the core of our innovations is a new conditional score function at the patch level, where the patch location in the original image is included as additional coordinate channels, while the patch size is randomized and diversified throughout training to encode the cross-region dependency at multiple scales. Sampling with our method is as easy as in the original diffusion model. Through Patch Diffusion, we could achieve $\mathbf{\ge 2\times}$ faster training, while maintaining comparable or better generation quality. Patch Diffusion meanwhile improves the performance of diffusion models trained on relatively small datasets, $e.g.$, as few as 5,000 images to train from scratch. We achieve state-of-the-art FID scores 1.77 on CelebA-64$\times$64 and 1.93 on AFHQv2-Wild-64$\times$64. We will share our code and pre-trained models soon.

Through prompting, large-scale pre-trained models have become more expressive and powerful, gaining significant attention in recent years. Though these big models have zero-shot capabilities, in general, labeled data are still required to adapt them to downstream tasks. To overcome this critical limitation, we propose an unsupervised fine-tuning framework to directly fine-tune the model or prompt on the unlabeled target data. We demonstrate how to apply our method to both language-augmented vision and masked-language models by aligning the discrete distributions extracted from the prompts and target data. To verify our approach's applicability, we conduct extensive experiments on image classification, sentiment analysis, and natural language inference tasks. Across 13 image-related tasks and 15 language-related ones, the proposed approach achieves consistent improvements over the baselines., Comment: ICML 2023; PyTorch code is available at https://github.com/korawat-tanwisuth/POUF

Tailoring outputs of large language models, such as ChatGPT, to specific user needs remains a challenge despite their impressive generation quality. In this paper, we propose a tri-agent generation pipeline consisting of a generator, an instructor, and an editor to enhance the customization of generated outputs. The generator produces an initial output, the user-specific instructor generates editing instructions, and the editor generates a revised output aligned with user preferences. The inference-only large language model (ChatGPT) serves as both the generator and the editor, while a smaller model acts as the user-specific instructor to guide the generation process toward user needs. The instructor is trained using editor-steered reinforcement learning, leveraging feedback from the large-scale editor model to optimize instruction generation. Experimental results on two abstractive summarization datasets demonstrate the effectiveness of our approach in generating outputs that better fulfill user expectations., Comment: 12 pages, work in progress

Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models. We also collect feedback and comparison data from GPT-4 to enable a comprehensive evaluation and reward model training. We make our data generated using GPT-4 as well as our codebase publicly available., Comment: 8 pages. Work in progress. Project page: https://instruction-tuning-with-gpt-4.github.io

We present Prompt Diffusion, a framework for enabling in-context learning in diffusion-based generative models. Given a pair of task-specific example images, such as depth from/to image and scribble from/to image, and a text guidance, our model automatically understands the underlying task and performs the same task on a new query image following the text guidance. To achieve this, we propose a vision-language prompt that can model a wide range of vision-language tasks and a diffusion model that takes it as input. The diffusion model is trained jointly over six different tasks using these prompts. The resulting Prompt Diffusion model is the first diffusion-based vision-language foundation model capable of in-context learning. It demonstrates high-quality in-context generation on the trained tasks and generalizes effectively to new, unseen vision tasks with their respective prompts. Our model also shows compelling text-guided image editing results. Our framework, with code publicly available at https://github.com/Zhendong-Wang/Prompt-Diffusion, aims to facilitate research into in-context learning for computer vision.

Content-Controllable Summarization generates summaries focused on the given controlling signals. Due to the lack of large-scale training corpora for the task, we propose a plug-and-play module RelAttn to adapt any general summarizers to the content-controllable summarization task. RelAttn first identifies the relevant content in the source documents, and then makes the model attend to the right context by directly steering the attention weight. We further apply an unsupervised online adaptive parameter searching algorithm to determine the degree of control in the zero-shot setting, while such parameters are learned in the few-shot setting. By applying the module to three backbone summarization models, experiments show that our method effectively improves all the summarizers, and outperforms the prefix-based method and a widely used plug-and-play model in both zero- and few-shot settings. Tellingly, more benefit is observed in the scenarios when more control is needed.

Dialogue summarization has recently garnered significant attention due to its wide range of applications. However, existing methods for summarizing dialogues have limitations because they do not take into account the inherent structure of dialogue and rely heavily on labeled data, which can lead to poor performance in new domains. In this work, we propose DIONYSUS (dynamic input optimization in pre-training for dialogue summarization), a pre-trained encoder-decoder model for summarizing dialogues in any new domain. To pre-train DIONYSUS, we create two pseudo summaries for each dialogue example: one is produced by a fine-tuned summarization model, and the other is a collection of dialogue turns that convey important information. We then choose one of these pseudo summaries based on the difference in information distribution across different types of dialogues. This selected pseudo summary serves as the objective for pre-training DIONYSUS using a self-supervised approach on a large dialogue corpus. Our experiments show that DIONYSUS outperforms existing methods on six datasets, as demonstrated by its ROUGE scores in zero-shot and few-shot settings., Accepted for ACL 2023

This paper presents Z-Code++, a new pre-trained language model optimized for abstractive text summarization. The model extends the state of the art encoder-decoder model using three techniques. First, we use a two-phase pre-training process to improve model's performance on low-resource summarization tasks. The model is first pre-trained using text corpora for language understanding, and then is continually pre-trained on summarization corpora for grounded text generation. Second, we replace self-attention layers in the encoder with disentangled attention layers, where each word is represented using two vectors that encode its content and position, respectively. Third, we use fusion-in-encoder, a simple yet effective method of encoding long sequences in a hierarchical manner. Z-Code++ creates new state of the art on 9 out of 13 text summarization tasks across 5 languages. Our model is parameter-efficient in that it outperforms the 600x larger PaLM-540B on XSum, and the finetuned 200x larger GPT3-175B on SAMSum. In zero-shot and few-shot settings, our model substantially outperforms the competing models., 16 pages, 3 figures. Accepted as long paper in main conference of ACL 2023

We introduce GODEL (Grounded Open Dialogue Language Model), a large pre-trained language model for dialog. In contrast with earlier models such as DialoGPT, GODEL leverages a new phase of grounded pre-training designed to better support adapting GODEL to a wide range of downstream dialog tasks that require information external to the current conversation (e.g., a database or document) to produce good responses. Experiments against an array of benchmarks that encompass task-oriented dialog, conversational QA, and grounded open-domain dialog show that GODEL outperforms state-of-the-art pre-trained dialog models in few-shot fine-tuning setups, in terms of both human and automatic evaluation. A novel feature of our evaluation methodology is the introduction of a notion of utility that assesses the usefulness of responses (extrinsic evaluation) in addition to their communicative features (intrinsic evaluation). We show that extrinsic evaluation offers improved inter-annotator agreement and correlation with automated metrics. Code and data processing scripts are publicly available.

Token-mixing multi-layer perceptron (MLP) models have shown competitive performance in computer vision tasks with a simple architecture and relatively small computational cost. Their success in maintaining computation efficiency is mainly attributed to avoiding the use of self-attention that is often computationally heavy, yet this is at the expense of not being able to mix tokens both globally and locally. In this paper, to exploit both global and local dependencies without self-attention, we present Mix-Shift-MLP (MS-MLP) which makes the size of the local receptive field used for mixing increase with respect to the amount of spatial shifting. In addition to conventional mixing and shifting techniques, MS-MLP mixes both neighboring and distant tokens from fine- to coarse-grained levels and then gathers them via a shifting operation. This directly contributes to the interactions between global and local tokens. Being simple to implement, MS-MLP achieves competitive performance in multiple vision benchmarks. For example, an MS-MLP with 85 million parameters achieves 83.8% top-1 classification accuracy on ImageNet-1K. Moreover, by combining MS-MLP with state-of-the-art Vision Transformers such as the Swin Transformer, we show MS-MLP achieves further improvements on three different model scales, e.g., by 0.5% on ImageNet-1K classification with Swin-B. The code is available at: https://github.com/JegZheng/MS-MLP.

Fine-tuning large language models for different tasks can be costly and inefficient, and even methods that reduce the number of tuned parameters still require full gradient-based optimization. We propose HyperTuning, a novel approach to model adaptation that uses a hypermodel to generate task-specific parameters for a fixed downstream model. We demonstrate a simple setup for hypertuning with HyperT5, a T5-based hypermodel that produces soft prefixes or LoRA parameters for a frozen T5 model from few-shot examples. We train HyperT5 in two stages: first, hyperpretraining with a modified conditional language modeling objective that trains a hypermodel to generate parameters; second, multi-task fine-tuning (MTF) on a large number of diverse language tasks. We evaluate HyperT5 on P3, MetaICL and Super-NaturalInstructions datasets, and show that it can effectively generate parameters for unseen tasks. Moreover, we show that using hypermodel-generated parameters as initializations for further parameter-efficient fine-tuning improves performance. HyperTuning can thus be a flexible and efficient way to leverage large language models for diverse downstream applications.

Large Transformer-based models have exhibited superior performance in various natural language processing and computer vision tasks. However, these models contain enormous amounts of parameters, which restrict their deployment to real-world applications. To reduce the model size, researchers prune these models based on the weights' importance scores. However, such scores are usually estimated on mini-batches during training, which incurs large variability/uncertainty due to mini-batch sampling and complicated training dynamics. As a result, some crucial weights could be pruned by commonly used pruning methods because of such uncertainty, which makes training unstable and hurts generalization. To resolve this issue, we propose PLATON, which captures the uncertainty of importance scores by upper confidence bound (UCB) of importance estimation. In particular, for the weights with low importance scores but high uncertainty, PLATON tends to retain them and explores their capacity. We conduct extensive experiments with several Transformer-based models on natural language understanding, question answering and image classification to validate the effectiveness of PLATON. Results demonstrate that PLATON manifests notable improvement under different sparsity levels. Our code is publicly available at https://github.com/QingruZhang/PLATON., Comment: Proceedings of the 39th International Conference on Machine Learning (ICML 2022)

Model ensemble is a popular approach to produce a low-variance and well-generalized model. However, it induces large memory and inference costs, which are often not affordable for real-world deployment. Existing work has resorted to sharing weights among models. However, when increasing the proportion of the shared weights, the resulting models tend to be similar, and the benefits of using model ensemble diminish. To retain ensemble benefits while maintaining a low memory cost, we propose a consistency-regularized ensemble learning approach based on perturbed models, named CAMERO. Specifically, we share the weights of bottom layers across all models and apply different perturbations to the hidden representations for different models, which can effectively promote the model diversity. Meanwhile, we apply a prediction consistency regularizer across the perturbed models to control the variance due to the model diversity. Our experiments using large language models demonstrate that CAMERO significantly improves the generalization performance of the ensemble model. Specifically, CAMERO outperforms the standard ensemble of 8 BERT-base models on the GLUE benchmark by 0.7 with a significantly smaller model size (114.2M vs. 880.6M)., Comment: Proceedings of ACL 2022

Employing a forward diffusion chain to gradually map the data to a noise distribution, diffusion-based generative models learn how to generate the data by inferring a reverse diffusion chain. However, this approach is slow and costly because it needs many forward and reverse steps. We propose a faster and cheaper approach that adds noise not until the data become pure random noise, but until they reach a hidden noisy-data distribution that we can confidently learn. Then, we use fewer reverse steps to generate data by starting from this hidden distribution that is made similar to the noisy data. We reveal that the proposed model can be cast as an adversarial auto-encoder empowered by both the diffusion process and a learnable implicit prior. Experimental results show even with a significantly smaller number of reverse diffusion steps, the proposed truncated diffusion probabilistic models can provide consistent improvements over the non-truncated ones in terms of performance in both unconditional and text-guided image generations.

Recent research has shown the existence of significant redundancy in large Transformer models. One can prune the redundant parameters without significantly sacrificing the generalization performance. However, we question whether the redundant parameters could have contributed more if they were properly trained. To answer this question, we propose a novel training strategy that encourages all parameters to be trained sufficiently. Specifically, we adaptively adjust the learning rate for each parameter according to its sensitivity, a robust gradient-based measure reflecting this parameter's contribution to the model performance. A parameter with low sensitivity is redundant, and we improve its fitting by increasing its learning rate. In contrast, a parameter with high sensitivity is well-trained, and we regularize it by decreasing its learning rate to prevent further overfitting. We conduct extensive experiments on natural language understanding, neural machine translation, and image classification to demonstrate the effectiveness of the proposed schedule. Analysis shows that the proposed schedule indeed reduces the redundancy and improves generalization performance., Comment: Proceedings of ICLR 2022

Tertiary dentin results from the interplay between the host defense and dental injury or infection. Modern endodontics aiming vital pulp treatment take the tertiary dentin formation as the interim step, with the final goal of a physiological pulp-dentin like tissue regeneration. Dental pulp stem cells have been nominated for contributing to differentiating into odontoblast-like cells who are responsible for reparative dentin formation. Understanding the original dentin formation mechanism provides us a blueprint while exploring the reparative dentin formation mechanism builds bridge to bonafide pulp-dentin tissue regeneration. Among all the regulators, growth factors have long been revealed under the spotlight. The insulin-like growth factor (IGF) family has been implicated in critical events of inducing dentin formation, which is essential for pulp treatment. The expression of IGF family members including IGF1, IGF1R, IGF2, and IGF2R has been well characterized in dental papilla cells, dental pulp stem cells, and periodontal ligament cells. Recent studies indicated IGF binding to the receptors activated pathways, including MAPK pathway, and AKT pathway, orchestrated proliferation, and differentiation, and finally, contributed to dentin formation. This review summarizes the role of IGF family in dentin formation during tooth development and tertiary dentin formation during dentin-pulp repair and sheds light on key parts of research for future treatment improvements.

Generative adversarial networks (GANs) are challenging to train stably, and a promising remedy of injecting instance noise into the discriminator input has not been very effective in practice. In this paper, we propose Diffusion-GAN, a novel GAN framework that leverages a forward diffusion chain to generate Gaussian-mixture distributed instance noise. Diffusion-GAN consists of three components, including an adaptive diffusion process, a diffusion timestep-dependent discriminator, and a generator. Both the observed and generated data are diffused by the same adaptive diffusion process. At each diffusion timestep, there is a different noise-to-data ratio and the timestep-dependent discriminator learns to distinguish the diffused real data from the diffused generated data. The generator learns from the discriminator's feedback by backpropagating through the forward diffusion chain, whose length is adaptively adjusted to balance the noise and data levels. We theoretically show that the discriminator's timestep-dependent strategy gives consistent and helpful guidance to the generator, enabling it to match the true data distribution. We demonstrate the advantages of Diffusion-GAN over strong GAN baselines on various datasets, showing that it can produce more realistic images with higher stability and data efficiency than state-of-the-art GANs., Comment: Project homepage: https://github.com/Zhendong-Wang/Diffusion-GAN

The input and output of most text generation tasks can be transformed to two sequences of tokens and they can be modeled using sequence-to-sequence learning modeling tools such as Transformers. These models are usually trained by maximizing the likelihood the output text sequence and assumes the input sequence and all gold preceding tokens are given during training, while during inference the model suffers from the exposure bias problem (i.e., it only has access to its previously predicted tokens rather gold tokens during beam search). In this paper, we propose MoCa ({\bf Mo}mentum {\bf Ca}libration) for text generation. MoCa is an online method that dynamically generates slowly evolving (but consistent) samples using a momentum moving average generator with beam search and MoCa learns to align its model scores of these samples with their actual qualities. Experiments on four text generation datasets (i.e., CNN/DailyMail, XSum, SAMSum and Gigaword) show MoCa consistently improves strong pre-trained transformers using vanilla fine-tuning and we achieve the state-of-the-art results on CNN/DailyMail and SAMSum datasets.

Layer-wise distillation is a powerful tool to compress large models (i.e. teacher models) into small ones (i.e., student models). The student distills knowledge from the teacher by mimicking the hidden representations of the teacher at every intermediate layer. However, layer-wise distillation is difficult. Since the student has a smaller model capacity than the teacher, it is often under-fitted. Furthermore, the hidden representations of the teacher contain redundant information that the student does not necessarily need for the target task's learning. To address these challenges, we propose a novel Task-aware layEr-wise Distillation (TED). TED designs task-aware filters to align the hidden representations of the student and the teacher at each layer. The filters select the knowledge that is useful for the target task from the hidden representations. As such, TED reduces the knowledge gap between the two models and helps the student to fit better on the target task. We evaluate TED in two scenarios: continual pre-training and fine-tuning. TED demonstrates significant and consistent improvements over existing distillation methods in both scenarios. Code is available at https://github.com/cliang1453/task-aware-distillation., Comment: Proceedings of ICML 2023

We review the EfficientQA competition from NeurIPS 2020. The competition focused on open-domain question answering (QA), where systems take natural language questions as input and return natural language answers. The aim of the competition was to build systems that can predict correct answers while also satisfying strict on-disk memory budgets. These memory budgets were designed to encourage contestants to explore the trade-off between storing retrieval corpora or the parameters of learned models. In this report, we describe the motivation and organization of the competition, review the best submissions, and analyze system predictions to inform a discussion of evaluation for open-domain QA., Comment: 26 pages; Published in Proceedings of Machine Learning Research (PMLR), NeurIPS 2020 Competition and Demonstration Track

Adversarial regularization can improve model generalization in many natural language processing tasks. However, conventional approaches are computationally expensive since they need to generate a perturbation for each sample in each epoch. We propose a new adversarial regularization method ARCH (adversarial regularization with caching), where perturbations are generated and cached once every several epochs. As caching all the perturbations imposes memory usage concerns, we adopt a K-nearest neighbors-based strategy to tackle this issue. The strategy only requires caching a small amount of perturbations, without introducing additional training time. We evaluate our proposed method on a set of neural machine translation and natural language understanding tasks. We observe that ARCH significantly eases the computational burden (saves up to 70% of computational time in comparison with conventional approaches). More surprisingly, by reducing the variance of stochastic gradients, ARCH produces a notably better (in most of the tasks) or comparable model generalization. Our code is available at https://github.com/SimiaoZuo/Caching-Adv., EMNLP 2021 (findings)

Additional file 1: Table S1. Association of platelet count on infection after multivariable adjustment.

This paper presents a new pre-trained language model, DeBERTaV3, which improves the original DeBERTa model by replacing mask language modeling (MLM) with replaced token detection (RTD), a more sample-efficient pre-training task. Our analysis shows that vanilla embedding sharing in ELECTRA hurts training efficiency and model performance. This is because the training losses of the discriminator and the generator pull token embeddings in different directions, creating the "tug-of-war" dynamics. We thus propose a new gradient-disentangled embedding sharing method that avoids the tug-of-war dynamics, improving both training efficiency and the quality of the pre-trained model. We have pre-trained DeBERTaV3 using the same settings as DeBERTa to demonstrate its exceptional performance on a wide range of downstream natural language understanding (NLU) tasks. Taking the GLUE benchmark with eight tasks as an example, the DeBERTaV3 Large model achieves a 91.37% average score, which is 1.37% over DeBERTa and 1.91% over ELECTRA, setting a new state-of-the-art (SOTA) among the models with a similar structure. Furthermore, we have pre-trained a multi-lingual model mDeBERTa and observed a larger improvement over strong baselines compared to English models. For example, the mDeBERTa Base achieves a 79.8% zero-shot cross-lingual accuracy on XNLI and a 3.6% improvement over XLM-R Base, creating a new SOTA on this benchmark. We have made our pre-trained models and inference code publicly available at https://github.com/microsoft/DeBERTa., Comment: 16 pages, 10 tables, 2 Figures. The DeBERTaV3 model significantly improves performance of the downstream NLU tasks over models with a similar structure, e.g. DeBERTaV3 large achieves 91.37% average GLUE score which is 1.37% over DeBERTa large. XSmall has only 22M backbone parameters, but significantly outperforms RoBERTa/XLNet-base. Paper is published as a conference paper at ICLR 2023

Adversarial regularization has been shown to improve the generalization performance of deep learning models in various natural language processing tasks. Existing works usually formulate the method as a zero-sum game, which is solved by alternating gradient descent/ascent algorithms. Such a formulation treats the adversarial and the defending players equally, which is undesirable because only the defending player contributes to the generalization performance. To address this issue, we propose Stackelberg Adversarial Regularization (SALT), which formulates adversarial regularization as a Stackelberg game. This formulation induces a competition between a leader and a follower, where the follower generates perturbations, and the leader trains the model subject to the perturbations. Different from conventional approaches, in SALT, the leader is in an advantageous position. When the leader moves, it recognizes the strategy of the follower and takes the anticipated follower's outcomes into consideration. Such a leader's advantage enables us to improve the model fitting to the unperturbed data. The leader's strategic information is captured by the Stackelberg gradient, which is obtained using an unrolling algorithm. Our experimental results on a set of machine translation and natural language understanding tasks show that SALT outperforms existing adversarial regularization baselines across all tasks. Our code is available at https://github.com/SimiaoZuo/Stackelberg-Adv., Comment: EMNLP 2021

Most of today's AI systems focus on using self-attention mechanisms and transformer architectures on large amounts of diverse data to achieve impressive performance gains. In this paper, we propose to augment the transformer architecture with an external attention mechanism to bring external knowledge and context to bear. By integrating external information into the prediction process, we hope to reduce the need for ever-larger models and increase the democratization of AI systems. We find that the proposed external attention mechanism can significantly improve the performance of existing AI systems, allowing practitioners to easily customize foundation AI models to many diverse downstream applications. In particular, we focus on the task of Commonsense Reasoning, demonstrating that the proposed external attention mechanism can augment existing transformer models and significantly improve the model's reasoning capabilities. The proposed system, Knowledgeable External Attention for commonsense Reasoning (KEAR), reaches human parity on the open CommonsenseQA research benchmark with an accuracy of 89.4\% in comparison to the human accuracy of 88.9\%., Comment: 8 pages, 1 figure, 7 tables

Current open-domain question answering systems often follow a Retriever-Reader architecture, where the retriever first retrieves relevant passages and the reader then reads the retrieved passages to form an answer. In this paper, we propose a simple and effective passage reranking method, named Reader-guIDEd Reranker (RIDER), which does not involve training and reranks the retrieved passages solely based on the top predictions of the reader before reranking. We show that RIDER, despite its simplicity, achieves 10 to 20 absolute gains in top-1 retrieval accuracy and 1 to 4 Exact Match (EM) gains without refining the retriever or reader. In addition, RIDER, without any training, outperforms state-of-the-art transformer-based supervised rerankers. Remarkably, RIDER achieves 48.3 EM on the Natural Questions dataset and 66.4 EM on the TriviaQA dataset when only 1,024 tokens (7.8 passages on average) are used as the reader input after passage reranking., Comment: Minor format updates. TLDR: Reranking retrieved passages by reader predictions can achieve 10~20 gains in top-1 retrieval accuracy and 1~4 gains in Exact Match (EM) without any training

Most of the policy evaluation algorithms are based on the theories of Bellman Expectation and Optimality Equation, which derive two popular approaches - Policy Iteration (PI) and Value Iteration (VI). However, multi-step bootstrapping is often at cross-purposes with and off-policy learning in PI-based methods due to the large variance of multi-step off-policy correction. In contrast, VI-based methods are naturally off-policy but subject to one-step learning.In this paper, we deduce a novel multi-step Bellman Optimality Equation by utilizing a latent structure of multi-step bootstrapping with the optimal value function. Via this new equation, we derive a new multi-step value iteration method that converges to the optimal value function with exponential contraction rate $\mathcal{O}(\gamma^n)$ but only linear computational complexity. Moreover, it can naturally derive a suite of multi-step off-policy algorithms that can safely utilize data collected by arbitrary policies without correction.Experiments reveal that the proposed methods are reliable, easy to implement and achieve state-of-the-art performance on a series of standard benchmark datasets.

Shuguang Wen,1– 3,* Weili Wang,4,* Ruimin Liu,3 Pengcheng He1 1Department of Hematology, The First Affiliated Hospital of Xi’an Jiaotong University, Xi’an 710061, People’s Republic of China; 2Department of Hematology, The First Affiliated Hospital of Zhengzhou University, Zhengzhou 450052, People’s Republic of China; 3Basic Medical College, Henan University, Kaifeng 475000, People’s Republic of China; 4State Key Laboratory of Radiation Medicine and Protection, School for Radiological and Interdisciplinary Sciences (RAD-X), Collaborative Innovation Center of Radiological Medicine of Jiangsu Higher Education Institutions, Soochow University, Suzhou 215123, People’s Republic of China*These authors contributed equally to this workCorrespondence: Pengcheng HeDepartment of Hematology, The First Affiliated Hospital of Xi’an Jiaotong University, No. 277 Yanta West Road, Xi’an 710061, People’s Republic of ChinaTel +86 189 9123 2609Email hepengcheng@xjtu.edu.cnBackground: Fluorescent metallic nanodots (NDs) have become a promising nanoprobe for a wide range of biomedical applications. Because Ag NDs have a high tendency to be oxidized, their synthesis and storage are a big challenge. Thus, the method for preparing stable Ag NDs is urgently needed. Surface modification and functionalization can enrich the capability of Ag NDs.Methods: In this work, fluorescent Ag NDs were synthesized in deoxygenated water by using porcine pancreatic α-amylase (PPA) as the stabilizing/capping agent. The absorption and fluorescence of PPA-protected Ag NDs (PPA@AgNDs) were measured with a spectrophotometer and a spectrofluorometer, respectively. The morphology of PPA@AgNDs was characterized by high-angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM). The biocompatibility of PPA@AgNDs was evaluated by tetrazolium (MTT)-based assay. PolyLys-Cys-SH (sequence: KKKKKKC) peptides were conjugated to PPA@AgNDs via heterobifunctional crosslinkers. PolyLys-Cys-linked PPA@AgNDs absorbed 5-aminolevulinic acid (ALA) by electrostatic interaction at physiological pH. The capability of tumor targeting was evaluated by intravenously injecting PPA@AgND-ALA into 4T1 breast cancer xenograft mouse models. Photodynamic therapy (PDT) against tumors was performed under 635 nm laser irradiation.Results: PPA@AgNDs emitted at 640 nm with quantum yield of 2.1%. The Ag NDs exhibited strong photostability over a long period and a fluorescence lifetime of 5.1 ns. PPA@AgNDs easily entered the cells to stain the nuclei, showing the capabilities of living cell imaging with negligible cytotoxicity. ALA-loaded PPA@AgNDs (PPA@AgND-ALA) presented the superiority of passive tumor targeting via the enhanced permeability and retention (EPR) effect. Tumors were visualized in the near-infrared (NIR) region with reduced background noise. ALA molecules released from PPA@AgND-ALA was converted into the photosensitizer (PS) of protoporphyrin IX (PpIX) intracellularly and intratumorally, which greatly improved the PDT efficacy.Conclusion: Our approach opens a new way to design a novel theranostic nanoplatform of PPA@AgND-ALA for effective tumor targeting and fluorescence image-guided PDT.Keywords: α-amylase, Ag nanodots, peptides, targeted imaging, photodynamic therapy

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., Comment: 13 pages, 9 tables, 2 figures

Stem xylem-specific hydraulic conductivity (KS) represents the potential for plant water transport normalized by xylem cross section, length, and driving force. Variation in KS has implications for plant transpiration and photosynthesis, growth and survival, and also the geographic distribution of species. Clarifying the global- scale patterns of KS and its major drivers is needed to achieve a better understanding of how plants adapt to different environmental conditions, particularly under climate change scenarios. Here, we compiled a xylem hydraulics dataset with 1,186 species- at-site combinations (975 woody species representing 146 families, from 199 sites worldwide), and investigated how KS varied with climatic variables, plant functional types, and biomes. Growing-season temperature and growing-season precipitation drove global variation in KS independently. Both the mean and the variation in KS were highest in the warm and wet tropical regions, and lower in cold and dry regions, such as tundra and desert biomes. Our results suggest that future warming and redistribution of seasonal precipitation may have a significant impact on species functional diversity, and is likely to be particularly important in regions becoming warmer or drier, such as high latitudes. This highlights an important role for KS in predicting shifts in community composition in the face of climate change.

In this work, we present X-SQL, a new network architecture for the problem of parsing natural language to SQL query. X-SQL proposes to enhance the structural schema representation with the contextual output from BERT-style pre-training model, and together with type information to learn a new schema representation for down-stream tasks. We evaluated X-SQL on the WikiSQL dataset and show its new state-of-the-art performance.

The learning rate warmup heuristic achieves remarkable success in stabilizing training, accelerating convergence and improving generalization for adaptive stochastic optimization algorithms like RMSprop and Adam. Here, we study its mechanism in details. Pursuing the theory behind warmup, we identify a problem of the adaptive learning rate (i.e., it has problematically large variance in the early stage), suggest warmup works as a variance reduction technique, and provide both empirical and theoretical evidence to verify our hypothesis. We further propose RAdam, a new variant of Adam, by introducing a term to rectify the variance of the adaptive learning rate. Extensive experimental results on image classification, language modeling, and neural machine translation verify our intuition and demonstrate the effectiveness and robustness of our proposed method. All implementations are available at: https://github.com/LiyuanLucasLiu/RAdam., Comment: ICLR 2020. Fix several typos in the previous version

This paper explores the use of knowledge distillation to improve a Multi-Task Deep Neural Network (MT-DNN) (Liu et al., 2019) for learning text representations across multiple natural language understanding tasks. Although ensemble learning can improve model performance, serving an ensemble of large DNNs such as MT-DNN can be prohibitively expensive. Here we apply the knowledge distillation method (Hinton et al., 2015) in the multi-task learning setting. For each task, we train an ensemble of different MT-DNNs (teacher) that outperforms any single model, and then train a single MT-DNN (student) via multi-task learning to \emph{distill} knowledge from these ensemble teachers. We show that the distilled MT-DNN significantly outperforms the original MT-DNN on 7 out of 9 GLUE tasks, pushing the GLUE benchmark (single model) to 83.7\% (1.5\% absolute improvement\footnote{ Based on the GLUE leaderboard at https://gluebenchmark.com/leaderboard as of April 1, 2019.}). The code and pre-trained models will be made publicly available at https://github.com/namisan/mt-dnn., Comment: 8 pages, 2 figures and 3 tables

Objective: To investigate the effect of Danshen injection combined with prostaglandin E1 and low-dose heparinon prevention of hepatic veno-occlusive disease (HVOD) after hematopoietic stem cell transplantation (HSCT). Methods: A total of 126 patients undergoing HSCT between January 2008 and June 2015, received a combined prophylactic regimen of intravenous Danshen 20-30 ml/day, prostaglandin E 120-30 μg/d, and subcutaneous injection of low-dose heparin (100 U/kg/day) for prevention of HVOD. The incidence of HVOD and adverse events associated with this prophylactic regimen were observed. Results: Among 126 patients, 65 received autologous peripheral blood stem cell transplant, 34 received HLAidentical sibling HSCT, six received HLA mismatched siblingHSCT, seven received HLA-matched unrelated HSCT, and 14 received HLA-mismatched unrelated HSCT. No adverse reaction or coagulation disorder associated with this prophylactic regimen was observed. Only one case with acute myeloid leukemia (CR2) developed HVOD seven days after receiving myeloablative chemotherapy with oral busulfan and cyclophosphamide followed by HLAmismatched unrelated peripheral blood stem cell transplantation. This patient died of HVOD. Conclusion: Danshen injection combined with prostaglandin E1 and low-dose heparin is a safe and effective regimen for the prevention of HVOD after HSCT.

More and more researches show that hypertensive vascular remodeling is closely related to the imbalance of immune system in recent years. IFN-γ is natural protein with the function of immune regulation and has resistance effect on vascular remodeling. However, the mechanism of IFN-γ is to be defined. This paper is to explore the mechanism of IFN-γ in regulating OPN/Th17 pathway. In this research, animal models of vascular collagen remodeling were established by inducing hypertensive mice with ANG II. There was no statistical significance when the systolic blood pressures and the percentages of wall thickness/lumen diameter in both groups of WT + AngII + IFN-γ and WT + PBS were compared (P=0.219>0.05, P=0.118>0.05). The concentration of serum precollagen-type I and III and their ratio in WT + AngII + IFN-γ group were decreased after the IFN-γ being given (P

The numerical simulation of a 3-D model of the internal flow field for a U-shaped slide valve has been analyzed in this paper. The impact of different opening degree on the pressure and velocity were analyzed. The results show that in the U-shaped groove, the pressure changes rapidly, and in the outlet oil chamber, the vortex was appeared. According to the expression of the flow coefficient, the flow characteristic was predicted by CFD, the conclusions show that with the increase of opening degree, the flow gradually increases, and the flow coefficient was not a stable value.

This article describes an online blackout restoration decision support system, which includes black start optimization strategy, calculation steps, the main function and structure of the software platform. By common information platform, this system automatically and efficiently collect real time power system information & data from EMS, enabling the yearly rolling revision of the black-start scheme, or online adjustment, and provide a black start analysis and decision making tools for dispatcher. The system integrate network data management and scientific computing, graphical modeling input and output tools, providing a flexible platform for power system dispatch department to develop a high-quality black-start scheme by detailed data analysis.