AI Technology & Industry Review
In the new paper Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding, a Google Brain research team presents Imagen, a text-to-image diffusion model that combines deep language understanding and photorealistic image generation capabilities to achieve a new state-of-the-art FID score of 7.27 on the COCO dataset.
In the new paper Tracing Knowledge in Language Models Back to the Training Data, a team from MIT CSAIL and Google Research proposes a benchmark for tracing language models’ assertions to the associated training data, aiming to establish a principled ground truth and mitigate high compute demands for large neural language model training.
In the new paper Large Language Models are Zero-Shot Reasoners, a research team from the University of Tokyo and Google Brain demonstrates that large language models (LLMs) can become good zero-shot reasoners through the addition of a simple prompt — “Let’s think step by step” — that elicits a step-by-step thinking process before each question is answered. Their Zero-shot-CoT model achieves huge performance gains compared to the zero-shot baseline.
In the new paper Unified Pretraining Framework for Document Understanding, an Adobe Research and Adobe Document Cloud team presents a unified pretraining framework for document understanding that enables cross-modal connections, relevant information highlighting in both visual and textual modalities, and cross-modal connections. UDoc achieves impressive performance on various downstream tasks.
In the new paper Training Compute-Optimal Large Language Models, a DeepMind research team posits that current large language models are significantly undertrained and, based on empirical outcomes of over 400 training runs, proposes three predictive approaches for optimally setting model size and training duration.
In the new paper Token Dropping for Efficient BERT Pretraining, a research team from Google, New York University, and the University of Maryland proposes a simple but effective “token dropping” technique that significantly reduces the pretraining cost of transformer models such as BERT without hurting performance on downstream fine-tuning tasks.
A team from Google Research and the Swiss AI Lab IDSIA proposes the Block-Recurrent Transformer, a novel long-sequence processing approach that has the same computation time and parameter count costs as a conventional transformer layer but achieves significant perplexity improvements in language modelling tasks over very long sequences.
A Google research team publishes guidelines for designing more practical and reliable sparse expert models. Their pretrained 269B sparse model achieves state-of-the-art results across many natural language processing (NLP) benchmarks.
A research team from Microsoft and NVIDIA leverages the NVIDIA Megatron-LM and Microsoft’s DeepSpeed to create an efficient and scalable 3D parallel system that combines data, pipeline, and tensor-slicing based parallelism, achieving superior zero-, one-, and few-shot learning accuracies and new state-of-the-art results on NLP benchmarks.
A research team from Sapienza University and OpenAI introduces an explanatory learning procedure that enables machines to understand existing explanations from symbolic sequences and create new explanations for unexplained phenomena, and further proposes Critical Rationalist Network (CRN) models for discovering explanations for novel phenomena.
In the new paper Towards Interactive Language Modeling, a research team from the University of Amsterdam and Meta AI Labs presents a road map detailing the steps to be taken towards interactive language modelling.
A Google Research team explores the effects of scaling both input length and model size at the same time with LongT5, a novel transformer architecture that achieves state-of-the-art performance on long-sequence tasks.
An OpenAI research team fine-tunes the GPT-3 pretrained language model to enable it to answer long-form questions by searching and navigating a text-based web browsing environment, achieving retrieval and synthesis improvements and reaching human-level long-form question-answering performance.
Peng Cheng Laboratory (PCL) and Baidu release PCL-BAIDU Wenxin, the world’s first knowledge-enhanced 100-billion-scale pretrained language model and the largest Chinese-language monolithic model with 260 billion parameters. PCL-BAIDU Wenxin achieves state-of-the-art results on more than 60 tasks and significantly advances more than 30 benchmarks for zero-shot and few-shot learning.
A research team from the University of Washington, Facebook AI Research and the Allen Institute for AI introduces Meta-training for InContext Learning (MetaICL), a new meta-training framework for few-shot learning where an LM is meta-trained to learn in-context — conditioning on training examples to recover the task and make predictions.
In the new paper LMSOC: An Approach for Socially Sensitive Pretraining, a Twitter Cortex research team proposes a simple but effective approach for learning both linguistically contextualized and socially sensitive representations in large-scale language models.
Facebook AI Research proposes NormFormer, an approach that improves pretraining perplexity and downstream task performance for both causal and masked language models, achieving GPT3-Large (1.3B) zero-shot performance 60 percent faster and improving fine-tuned GLUE performance by 1.9 percent.
A research team from the University of Southern California and Google proposes TOME, a “mention memory” approach to factual knowledge extraction for NLU tasks. A transformer model with attention over a semi-parametric representation of the entire Wikipedia text corpus, TOME can extract information without supervision and achieves strong performance on multiple open-domain question answering benchmarks.
On October 11, Microsoft introduced the largest and “the most powerful monolithic transformer language model” trained to date, a 530 billion parameter GPT-3-style generative language model.
In the paper Fine-Tuned Transformers Show Clusters of Similar Representations Across Layers, a research team from New York University and the University of North Carolina at Chapel Hill uses centered kernel alignment (CKA) to measure the similarity of representations across layers and explore how fine-tuning changes transformers’ learned representations.
A Google AI research team explores zero-label learning (training with synthetic data only) in natural language processing, and introduces Unsupervised Data Generation (UDG), a training data creation procedure designed to synthesize high-quality training data without human annotations.
MIT researchers present an automated, objective and transparent data-driven method for measuring media bias. The study analyses roughly a million articles from about a hundred newspapers for bias on various news topics, maps the newspapers into a two-dimensional media bias landscape, and shows that the data-driven results agree well with human-judgement classifications.
A LinkedIn research team evaluates deep natural language processing (NLP) on various representative search engine tasks to provide insights for the development of industry search engines.
An Apple research team explores multiple architectures and training procedures to develop a novel multi-speaker and multi-lingual neural TTS system. The study combines speech from 30 speakers from 15 locales in 8 languages, and demonstrates that for the vast majority of voices, such multi-lingual and multi-speaker models can yield better quality than single speaker models.
A Huawei research team proposes a topic-based personalized ranking model (TPRM) that integrates pretrained contextualized term representations and user profiles constructed by a topic model to tailor the output ranking list.
A Google Research team explores the design space of Transformer models in an effort to enable deep learning architectures to solve compositional tasks. The proposed approach provides models with inductive biases via design decisions that significantly impact compositional generalization, and achieves state-of-the-art results on the COGS and PCFG composition benchmarks.
On August 5, WeChat AI and Beijing Jiaotong University system developers released the paper WeChat Neural Machine Translation Systems for WMT21, revealing the architecture of their novel neural machine translation (NMT) system and the strategies they adopted to achieve impressive performance in the WMT21 competition.
A Google Research team draws inspiration from two numerical analysis methods — Hierarchical Matrix (H-Matrix) and Multigrid — to address the quadratic complexity problem of attention mechanisms in transformer architectures, proposing a hierarchical attention scheme that has linear complexity in run time and memory.
A research team from the University of Melbourne, Facebook AI, and Twitter Cortex proposes a black-box test method for assessing and debugging the numerical translation of neural machine translation systems in a systematic manner. The approach reveals novel types of errors that are general across multiple state-of-the-art translation systems.
A Google Research team proposes Wordcraft, a text editor with a built-in AI-powered creative writing assistant. Wordcraft uses few-shot learning and the natural affordances of conversation to support a variety of user interactions; and can help with story planning, writing and editing.
A research team from OpenAI proposes Codex, a specialized GPT model fine-tuned on publicly available code from GitHub that can produce functionally correct Python code bodies from natural language docstrings and could excel at a variety of coding tasks.
A research team from ByteDance AI Lab, University of Wisconsin–Madison and Nanjing University wins the ACL 2021 best paper award. Their proposed Vocabulary Learning via Optimal Transport (VOLT) approach leverages optimal transport to automatically find an optimal vocabulary without trial training.
A Google Research team proposes MergeDistill, a framework for merging pretrained teacher LMs from multiple monolingual/multilingual LMs into a single multilingual task-agnostic student LM to leverage the capabilities of the powerful language-specific LMs while still being multilingual and enabling positive language transfer.
A research team from Google proposes ByT5 architecture, a competitive token-free pretrained byte-to-byte transformer that can be straightforwardly adapted to process byte sequences without adding excessive computational cost.
A research team from UC Davis, Microsoft Research and Johns Hopkins University extends work on training massive amounts of linguistic data to reveal the grammatical structures in their representations to the domain of mathematical reasoning, showing that both the standard transformer and the TP-Transformer can compose the meanings of mathematical symbols based on their structured relationships.
A research team from Microsoft Research and Peking University peeps into pretrained transformers and investigates how factual knowledge is stored, proposing a method to identify “knowledge neurons,” which can be utilized to explicitly update and erase facts.
A research team from McGill University, Mila – Quebec AI Institute and Facebook AI proposes novel metrics and perturbation functions to detect, quantify and compare trade-offs between robustness and faithfulness in NMT systems, both on the corpus level and with particular examples.
An IBM research team proposes four multilingual adversarial attack strategies and attacks seven languages in a zero-shot setting on large multilingual pretrained language models (e.g. MBERT), reducing average performance by up to 85.6 percent.
A Yann LeCun team proposes dictionary learning to provide detailed visualizations of transformer representations and insights into semantic structures such as word-level disambiguation, sentence-level pattern formation, and long-range dependency captured by transformers.
The Beijing Academy of Artificial Intelligence (BAAI) releases Wu Dao 1.0, China’s first large-scale pretraining model.