{"id":900798,"date":"2022-11-29T09:00:43","date_gmt":"2022-11-29T17:00:43","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?p=900798"},"modified":"2022-12-05T09:30:18","modified_gmt":"2022-12-05T17:30:18","slug":"research-focus-week-of-november-28-2022","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/research\/blog\/research-focus-week-of-november-28-2022\/","title":{"rendered":"Research Focus: Week of November 28, 2022"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"193\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-1024x193.jpg\" alt=\"Microsoft Research - Research Focus 05\nWeek of November 28th, 2022\" class=\"wp-image-901104\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-1024x193.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-300x57.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-768x145.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-1536x290.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-2048x386.jpg 2048w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x264_Research_focus5_blog_hero-240x45.jpg 240w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<figure class=\"wp-block-pullquote\"><blockquote><p><em>This special edition of Research Focus highlights some of the <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/event\/neurips-2022\/publications\/\" target=\"_blank\" rel=\"noreferrer noopener\">100+ papers from Microsoft Research<\/a> that were accepted for publication at NeurIPS&nbsp;2022 \u2013 the thirty-sixth annual Conference on Neural Information Processing Systems.<\/em><\/p><\/blockquote><\/figure>\n\n\n<aside id=accordion-5873e230-e802-4e65-b82f-52d8a61b66fc class=\"msr-table-of-contents-block accordion mb-5 pb-0\" data-bi-aN=\"table-of-contents\">\n\t<button class=\"btn btn-collapse bg-gray-100 mb-0 display-flex justify-content-between\" type=\"button\" data-mount=\"collapse\" data-target=\"#accordion-collapse-5873e230-e802-4e65-b82f-52d8a61b66fc\" aria-expanded=\"true\" aria-controls=\"accordion-collapse-5873e230-e802-4e65-b82f-52d8a61b66fc\">\n\t\t<span class=\"msr-table-of-contents-block__label subtitle\">In this article<\/span>\n\t\t<span class=\"msr-table-of-contents-block__current mr-4 text-gray-600 font-weight-normal\" aria-hidden=\"true\"><\/span>\n\t<\/button>\n\t<div id=\"accordion-collapse-5873e230-e802-4e65-b82f-52d8a61b66fc\" class=\"msr-table-of-contents-block__collapse-wrapper collapse show\" data-parent=\"#accordion-5873e230-e802-4e65-b82f-52d8a61b66fc\">\n\t\t<div class=\"accordion-body bg-gray-100 border-top pt-4\">\n\t\t\t<ol class=\"msr-table-of-contents-block__list\">\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">Few-shot Task-agnostic Neural Architecture Search for Distilling Large Language Models<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">Neuron with steady response leads to better generalization<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">Long-form video-language pre-training with multimodal temporal contrastive learning<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">Microsoft Research Causality and ML team features multiple papers and workshops at NeurIPS 2022<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">New research on generative models<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">Award Winner: A Neural Corpus Indexer for Document Retrieval<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t\t\t<li class=\"msr-table-of-contents-block__list-item\">\n\t\t\t\t\t\t<a href=\"\" class=\"msr-table-of-contents-block__list-item-link\">Microsoft Research career opportunities \u2013 come join us!<\/a>\n\t\t\t\t\t<\/li>\n\t\t\t\t\t\t\t<\/ul>\n\t\t<\/div>\n\t<\/div>\n\t<span class=\"msr-table-of-contents-block__progress-bar\"><\/span>\n<\/aside>\n\n\n\n<h2 id=\"few-shot-task-agnostic-neural-architecture-search-for-distilling-large-language-models\">Few-shot Task-agnostic Neural Architecture Search for Distilling Large Language Models<\/h2>\n\n\n\n<p><em>Dongkuan Xu, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/submukhe\/\" target=\"_blank\" rel=\"noreferrer noopener\">Subhabrata Mukherjee<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/xiaodl\/\" target=\"_blank\" rel=\"noreferrer noopener\">Xiaodong Liu<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/dedey\/\" target=\"_blank\" rel=\"noreferrer noopener\">Debadeepta Dey<\/a>, Wenhui Wang, Xiang Zhang, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/hassanam\/\" target=\"_blank\" rel=\"noreferrer noopener\">Ahmed Hassan Awadallah<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jfgao\/\" target=\"_blank\" rel=\"noreferrer noopener\">Jianfeng Gao<\/a><\/em><\/p>\n\n\n\n<p>Knowledge distillation (KD) is effective in compressing large pre-trained language models, where we train a small student model to mimic the output distribution of a large teacher model (e.g., BERT, GPT-X). KD relies on hand-designed student model architectures that require several trials and pre-specified compression rates. In our paper, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/autodistil-few-shot-task-agnostic-neural-architecture-search-for-distilling-large-language-models\/\" target=\"_blank\" rel=\"noreferrer noopener\">Few-shot Task-agnostic Neural Architecture Search for Distilling Large Language Models<\/a>, we discuss AutoDistil, a new technique pioneered by Microsoft Research that leverages advances in KD and neural architecture search (NAS) to automatically generate a suite of compressed models with variable computational cost (e.g., varying sizes, FLOPs and latency). NAS for distillation addresses customization challenges of hand-engineering compressed model architectures for diverse deployment environments having variable resource constraints with an automated framework. AutoDistil-generated compressed models obtain up to 41x reduction in FLOPs with limited regression in task performance and 6x FLOPs reduction with parity in performance with large teacher model. Given any state-of-the-art compressed model, AutoDistil finds a better compressed variant with better trade-off in task performance vs. computational cost during inference.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-1 wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button is-style-outline is-style-outline--9b67668a5940383bc007af40a79d102c\"><a data-bi-type=\"button\" class=\"wp-block-button__link\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/autodistil-few-shot-task-agnostic-neural-architecture-search-for-distilling-large-language-models\/\" target=\"_blank\" rel=\"noreferrer noopener\">Read the paper<\/a><\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-dots\"\/>\n\n\n\n<h2 id=\"neuron-with-steady-response-leads-to-better-generalization\">Neuron with steady response leads to better generalization<\/h2>\n\n\n\n<p><em>Qiang Fu, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/ludu\/\" target=\"_blank\" rel=\"noreferrer noopener\">Lun Du<\/a>, Haitao Mao, Xu Chen, Wei Fang, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/shihan\/\" target=\"_blank\" rel=\"noreferrer noopener\">Shi Han<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/dongmeiz\/\" target=\"_blank\" rel=\"noreferrer noopener\">Dongmei Zhang<\/a><\/em><\/p>\n\n\n\n<p>Improving models\u2019 ability to generalize is one of the most important research problems in machine learning. Deep neural networks with diverse architectures have been invented and widely applied to various domains and tasks. Our goal was to study and identify the fundamental properties commonly shared by different kinds of deep neural networks, and then design a generic technique applicable for all of them to improve their generalization.<\/p>\n\n\n\n<p>In this paper, from the neural level granularity, we study the characteristics of individual neurons\u2019 response during the training dynamics. We find that keeping the response of activated neurons stable for the same class helps improve models\u2019 ability to generalize. This is a new regularization perspective based on the neuron-level class-dependent response distribution. Meanwhile, we observed that the traditional vanilla model usually lacks good steadiness of intra-class response. Based on these observations, we designed a generic regularization method, Neuron Steadiness Regularization (NSR), to reduce large intra-class neuron response variance. NSR is computationally efficient and applicable to various architectures and tasks. Significant improvements are obtained on extensive experiments with multiple types of datasets and various network architectures. We will continue the research for improving the model generalization ability.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-2 wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button is-style-outline is-style-outline--23234f61a1df9fd30be0009006c7192e\"><a data-bi-type=\"button\" class=\"wp-block-button__link\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/neuron-with-steady-response-leads-to-better-generalization\/\" target=\"_blank\" rel=\"noreferrer noopener\">Read the paper<\/a><\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-dots\"\/>\n\n\n\n<h2 id=\"long-form-video-language-pre-training-with-multimodal-temporal-contrastive-learning\">Long-form video-language pre-training with multimodal temporal contrastive learning<\/h2>\n\n\n\n<p><em>Yuchong Sun, Hongwei Xue, Ruihua Song, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/libei\/\" target=\"_blank\" rel=\"noreferrer noopener\">Bei Liu<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/huayan\/\" target=\"_blank\" rel=\"noreferrer noopener\">Huan Yang<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jianf\/\" target=\"_blank\" rel=\"noreferrer noopener\">Jianlong Fu<\/a><\/em><\/p>\n\n\n\n<p>Huge numbers of videos on diverse topics and of various lengths are shared on social media. Analyzing and understanding these videos is an important but challenging problem. Previous work on action and scene recognition has been limited to certain labels, while neglecting the rich semantic and dynamic information in other videos. Inspired by the cross-modal pre-training paradigm in image-language domain (e.g., CLIP, Florence), researchers have explored video-language joint pre-training, which mainly use short-form videos (e.g., < 30 seconds). Long-form video and language pre-training have not been well studied yet, though long-form videos contain much richer and more complex semantic contents in real scenarios.<\/p>\n\n\n\n<p>In this research, we propose a <strong>L<\/strong>ong-<strong>F<\/strong>orm <strong>VI<\/strong>deo-<strong>LA<\/strong>nguage pre-training model (LF-VILA) to explore long-form video representation learning, and train it on a long-form video-language dataset (LF-VILA-8M) on the basis of our new collected video-language dataset (<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/advancing-high-resolution-video-language-representation-with-large-scale-video-transcriptions\/\">HD-VILA-100M<\/a>). We then design a <strong>M<\/strong>ultimodal <strong>T<\/strong>emporal <strong>C<\/strong>ontrastive (MTC) loss to capture the temporal relation between video clips and single sentences. We also propose the <strong>H<\/strong>ierarchical <strong>T<\/strong>emporal <strong>W<\/strong>indow <strong>A<\/strong>ttention (HTWA) mechanism on video encoder to reduce the training time by one-third. Our model achieves significant improvements on nine benchmarks, including paragraph-to-video retrieval, long-form video question-answering, and action recognition tasks. In the future, we will explore using it for broader scenarios, such as ego-centric video understanding.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-3 wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button is-style-outline is-style-outline--51891e8009c7c9ff25abcc975c564db6\"><a data-bi-type=\"button\" class=\"wp-block-button__link\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/advancing-high-resolution-video-language-representation-with-large-scale-video-transcriptions\/\" target=\"_blank\" rel=\"noreferrer noopener\">Read the paper<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-fill-github\"><a data-bi-type=\"button\" class=\"wp-block-button__link\" href=\"https:\/\/github.com\/microsoft\/XPretrain\" target=\"_blank\" rel=\"noreferrer noopener\">Download<\/a><\/div>\n\n\n\n<div class=\"wp-block-button is-style-outline is-style-outline--12ccdd7cddaeefb5401d66099c3c990b\"><a data-bi-type=\"button\" class=\"wp-block-button__link\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/long-form-video-language-pre-training-with-multimodal-temporal-contrastive-learning\/\" target=\"_blank\" rel=\"noreferrer noopener\">Read the paper<\/a><\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-dots\"\/>\n\n\n\n<h2 id=\"microsoft-research-causality-and-ml-team-features-multiple-papers-and-workshops-at-neurips-2022\">Microsoft Research Causality and ML team features multiple papers and workshops at NeurIPS 2022<\/h2>\n\n\n\n<p><em>Parikshit Bansal, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/ranveer\/\" target=\"_blank\" rel=\"noreferrer noopener\">Ranveer Chandra<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/eldillon\/\" target=\"_blank\" rel=\"noreferrer noopener\">Eleanor Dillon<\/a>, Saloni Dash, Rui Ding, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/daedge\/\" target=\"_blank\" rel=\"noreferrer noopener\">Darren Edge<\/a>, Adam Foster, Wenbo Gong, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/shihan\/\" target=\"_blank\" rel=\"noreferrer noopener\">Shi Han<\/a>, Agrin Hilmkil, Joel Jennings, Jian Jiao, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/emrek\/\" target=\"_blank\" rel=\"noreferrer noopener\">Emre K\u0131c\u0131man<\/a>, Hua Li, Chao Ma, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/saramalvar\/\" target=\"_blank\" rel=\"noreferrer noopener\">Sara Malvar<\/a>, Robert Ness, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/npawlowski\/\" target=\"_blank\" rel=\"noreferrer noopener\">Nick Pawlowski<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/yprabhu\/\" target=\"_blank\" rel=\"noreferrer noopener\">Yashoteja Prabhu<\/a>, Eduardo Rodrigues, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/amshar\/\" target=\"_blank\" rel=\"noreferrer noopener\">Amit Sharma<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/swatisharma\/\" target=\"_blank\" rel=\"noreferrer noopener\">Swati Sharma<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/chezha\/\" target=\"_blank\" rel=\"noreferrer noopener\">Cheng Zhang<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/dongmeiz\/\" target=\"_blank\" rel=\"noreferrer noopener\">Dongmei Zhang<\/a><\/em><\/p>\n\n\n\n<p>Identifying causal effects is an integral part of scientific inquiry, helping us to understand everything from educational outcomes to the effects of social policies to risk factors for diseases. Questions of cause-and-effect are also critical for the design and data-driven improvement and evaluation of business and technological systems we build today. The intersection of causal analysis and machine learning is driving rapid advances. Microsoft researchers are excited to be presenting three papers at NeurIPS, along with workshops on new methods and their applications. This includes work improving deep methods for causal discovery, applying causal insights to improve responsible language models, and improving soil carbon modeling with causal approaches. To accelerate research and broaden adoption of the latest causal methods, Microsoft researchers are co-organizing the <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.cml-4-impact.vanderschaar-lab.com\/\" target=\"_blank\" rel=\"noreferrer noopener\">Workshop on Causality for Real-world Impact<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> and releasing new no-code interactive <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/showwhy\" target=\"_blank\" rel=\"noreferrer noopener\">ShowWhy<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> tools for causal discovery and analysis. We encourage NeurIPS attendees to learn more via the links below or stop by the Microsoft booth for demos and talks.<\/p>\n\n\n\n<h3 id=\"main-conference-papers\">Main conference papers<\/h3>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-1 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Publication<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/simultaneous-missing-value-imputation-and-structure-learning-with-groups\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Simultaneous Missing Value Imputation and Structure Learning with Groups\" data-bi-aN=\"citation\" data-bi-cN=\"Simultaneous Missing Value Imputation and Structure Learning with Groups\">\n\t\t\t\tSimultaneous Missing Value Imputation and Structure Learning with Groups&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Publication<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/probing-classifiers-are-unreliable-for-concept-removal-and-detection\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Probing Classifiers are Unreliable for Concept Removal and Detection\" data-bi-aN=\"citation\" data-bi-cN=\"Probing Classifiers are Unreliable for Concept Removal and Detection\">\n\t\t\t\tProbing Classifiers are Unreliable for Concept Removal and Detection&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Publication<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/ml4s-learning-causal-skeleton-from-vicinal-graphs\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"ML4S: Learning Causal Skeleton from Vicinal Graphs\" data-bi-aN=\"citation\" data-bi-cN=\"ML4S: Learning Causal Skeleton from Vicinal Graphs\">\n\t\t\t\tML4S: Learning Causal Skeleton from Vicinal Graphs&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<h3 id=\"workshop-papers\">Workshop papers<\/h3>\n\n\n\n<p><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.cml-4-impact.vanderschaar-lab.com\/\" target=\"_blank\" rel=\"noreferrer noopener\"><strong>Workshop on Causality for Real-world Impact<\/strong><span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/a-causal-ai-suite-for-decision-making\/\">A Causal AI Suite for Decision-Making<\/a><\/li><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/the-counterfactual-shapley-value-attributing-change-in-system-metrics\/\" target=\"_blank\" rel=\"noreferrer noopener\">The Counterfactual-Shapley Value: Attributing Change in System Metrics<\/a><\/li><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/counterfactual-generation-under-confounding\/\" target=\"_blank\" rel=\"noreferrer noopener\">Counterfactual Generation Under Confounding<\/a><\/li><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/deep-end-to-end-causal-inference-2\/\" target=\"_blank\" rel=\"noreferrer noopener\">Deep End-to-end Causal Inference<\/a><\/li><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/rhino-deep-causal-temporal-relationship-learning-with-history-dependent-noise\/\" target=\"_blank\" rel=\"noreferrer noopener\">Rhino: Deep Causal Temporal Relationship Learning with history-dependent noise<\/a><\/li><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/causal-reasoning-in-the-presence-of-latent-confounders-via-neural-admg-learning\/\" target=\"_blank\" rel=\"noreferrer noopener\">Causal Reasoning in the Presence of Latent Confounders via Neural ADMG Learning<\/a><\/li><\/ul>\n\n\n\n<p><strong><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.climatechange.ai\/events\/neurips2022\" target=\"_blank\" rel=\"noreferrer noopener\">Workshop on Tackling Climate Change with Machine Learning<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/causal-modeling-of-soil-processes-for-improved-generalization\/\" target=\"_blank\" rel=\"noreferrer noopener\">Causal Modeling of Soil Processes for Improved Generalization<\/a><\/li><\/ul>\n\n\n\n<p><strong><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/sites.google.com\/view\/distshift2022\/home?pli=1\" target=\"_blank\" rel=\"noreferrer noopener\">Workshop on Distribution Shifts<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/using-interventions-to-improve-out-of-distribution-generalization-of-text-matching-recommendation-systems\/\" target=\"_blank\" rel=\"noreferrer noopener\">Using Interventions to Improve Out-of-Distribution Generalization of Text-Matching Recommendation Systems<\/a><\/li><\/ul>\n\n\n\n<p><strong><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/sites.google.com\/view\/icbinb-2022\/\" target=\"_blank\" rel=\"noreferrer noopener\">Workshop on Understanding Deep Learning Through Empirical Falsification (&#8220;I can&#8217;t believe it&#8217;s not better&#8221;)<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/strong><br><em>We\u2019ll be participating in the panel.<\/em><\/p>\n\n\n\n<div class=\"wp-block-group is-layout-flow wp-block-group-is-layout-flow\">\n<h4 id=\"causal-ai-software-resources\">Causal AI Software Resources<\/h4>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-2 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Download<\/span>\n\t\t\t<a href=\"https:\/\/github.com\/microsoft\/showwhy\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Causal No-Code Tools (ShowWhy)\" data-bi-aN=\"citation\" data-bi-cN=\"Causal No-Code Tools (ShowWhy)\">\n\t\t\t\tCausal No-Code Tools (ShowWhy)&nbsp;<span class=\"glyph-append glyph-append-share glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Video<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/video\/introduction-to-showwhy-user-interfaces-for-causal-decision-making\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Introduction to ShowWhy, user interfaces for causal decision making\" data-bi-aN=\"citation\" data-bi-cN=\"Introduction to ShowWhy, user interfaces for causal decision making\">\n\t\t\t\tIntroduction to ShowWhy, user interfaces for causal decision making&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-3 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Download<\/span>\n\t\t\t<a href=\"https:\/\/www.pywhy.org\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"An Open Source Ecosystem for Causal Machine Learning (PyWhy)\" data-bi-aN=\"citation\" data-bi-cN=\"An Open Source Ecosystem for Causal Machine Learning (PyWhy)\">\n\t\t\t\tAn Open Source Ecosystem for Causal Machine Learning (PyWhy)&nbsp;<span class=\"glyph-append glyph-append-share glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n\n\n\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Download<\/span>\n\t\t\t<a href=\"https:\/\/github.com\/microsoft\/causica\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Deep Causal Learning frameworks (Causica)\" data-bi-aN=\"citation\" data-bi-cN=\"Deep Causal Learning frameworks (Causica)\">\n\t\t\t\tDeep Causal Learning frameworks (Causica)&nbsp;<span class=\"glyph-append glyph-append-share glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\">\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Video<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/video\/causal-ai-for-decision-making\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Causal AI for Decision Making\" data-bi-aN=\"citation\" data-bi-cN=\"Causal AI for Decision Making\">\n\t\t\t\tCausal AI for Decision Making&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-dots\"\/>\n\n\n\n<h2 id=\"new-research-on-generative-models\">New research on generative models<\/h2>\n\n\n\n<p>Two papers covering new research on generative models will be presented at NeurIPS 2022.<\/p>\n\n\n\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Publication<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/operationalizing-specifications-in-addition-to-test-sets-for-evaluating-constrained-generative-models\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Operationalizing Specifications, In Addition to Test Sets for Evaluating Constrained Generative Models\" data-bi-aN=\"citation\" data-bi-cN=\"Operationalizing Specifications, In Addition to Test Sets for Evaluating Constrained Generative Models\">\n\t\t\t\tOperationalizing Specifications, In Addition to Test Sets for Evaluating Constrained Generative Models&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n\n\n\n<p><em><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/viraunak\/\" target=\"_blank\" rel=\"noreferrer noopener\">Vikas Raunak<\/a>, Matt Post, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/arulm\/\" target=\"_blank\" rel=\"noreferrer noopener\">Arul Menezes<\/a><\/em><\/p>\n\n\n\n<p>The first paper,\u202f<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/www.dropbox.com\/s\/vd0x2slkvmjkuq5\/Operationalizing_Specifications.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">Operationalizing Specifications, In Addition to Test Sets for Evaluating Constrained Generative Models<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, presents recommendations on the evaluation of state-of-the-art generative models for constrained generation tasks. The progress on generative models has been rapid in recent years. These large-scale models have had three impacts: 1) The fluency of generation in both language and vision modalities has rendered common average-case evaluation metrics much less useful in diagnosing system errors; 2) The same substrate models now form the basis of a number of applications, driven both by the utility of their representations as well as phenomena such as in-context learning, which raise the abstraction level of interacting with such models; 3) The user expectations around these models have made the technical challenge of out-of-domain generalization much less excusable in practice. Subsequently, our evaluation methodologies haven\u2019t adapted to these changes. More concretely, while the associated utility and methods of interacting with generative models have expanded, a similar expansion has not been observed in their evaluation practices. In this paper, we argue that the scale of generative models could be exploited to raise the abstraction level at which evaluation itself is conducted and provide recommendations for the same. Our recommendations are based on leveraging specifications as a powerful instrument to evaluate generation quality and are readily applicable to a variety of tasks.&nbsp;<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Venue: <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/humaneval-workshop.github.io\/\" target=\"_blank\" rel=\"noreferrer noopener\">Workshop on Human Evaluation of Generative Models, 36th Conference on Neural Information Processing Systems<span class=\"sr-only\"> (opens in new tab)<\/span><\/a><\/li><\/ul>\n\n\n\n<div class=\"annotations \" data-bi-aN=\"citation\">\n\t<ul class=\"annotations__list card depth-16 bg-body p-4 \">\n\t\t<li class=\"annotations__list-item\">\n\t\t\t\t\t\t<span class=\"annotations__type d-block text-uppercase font-weight-semibold text-neutral-300 small\">Publication<\/span>\n\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/rank-one-editing-of-encoder-decoder-models\/\" target=\"_self\" class=\"annotations__link font-weight-semibold text-decoration-none\" data-bi-type=\"annotated-link\" aria-label=\"Rank-One Editing of Encoder-Decoder Models\" data-bi-aN=\"citation\" data-bi-cN=\"Rank-One Editing of Encoder-Decoder Models\">\n\t\t\t\tRank-One Editing of Encoder-Decoder Models&nbsp;<span class=\"glyph-append glyph-append-chevron-right glyph-append-xsmall\"><\/span>\n\t\t\t<\/a>\n\t\t\t\t\t<\/li>\n\t<\/ul>\n<\/div>\n\n\n\n<p><em><em><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/viraunak\/\" target=\"_blank\" rel=\"noreferrer noopener\">Vikas Raunak<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/arulm\/\" target=\"_blank\" rel=\"noreferrer noopener\">Arul Menezes<\/a><\/em><\/em><\/p>\n\n\n\n<p>The second paper is\u202f<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/internlp.github.io\/documents\/2022\/papers\/19.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">Rank-One Editing of Encoder-Decoder Models.<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> Here, we look at large sequence-to-sequence models for tasks such as neural machine translation (NMT), which are usually trained over hundreds of millions of samples. However, training is just the origin of a model&#8217;s life-cycle. Real-world deployments of models require further behavioral adaptations as new requirements emerge or shortcomings become known. Typically, in the space of model behaviors, behavior deletion requests are addressed through model retrainings, whereas model finetuning is done to address behavior addition requests. Both procedures are instances of data-based model intervention. In this work, we present a preliminary study investigating rank-one editing as a direct intervention method for behavior deletion requests in encoder-decoder transformer models. We propose four editing tasks for NMT and show that the proposed editing algorithm achieves high efficacy, while requiring only a single instance of positive example to fix an erroneous (negative) model behavior. This research therefore explores a path towards fixing the deleterious behaviors of encoder-decoder models for tasks such as translation, making them safer and more reliable without investing in a huge computational budget.&nbsp;<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Venue:&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/internlp.github.io\/index.html\" target=\"_blank\" rel=\"noreferrer noopener\">The Second Workshop On Interactive Learning For Natural Language Processing<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> (InterNLP 2022)<\/li><\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-dots\"\/>\n\n\n\n<h2 id=\"award-winner-a-neural-corpus-indexer-for-document-retrieval\">Award Winner: A Neural Corpus Indexer for Document Retrieval<\/h2>\n\n\n\n<p><em>Yujing Wang, Yingyan Hou, Haonan Wang, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/zimiao\/\" target=\"_blank\" rel=\"noreferrer noopener\">Ziming Miao<\/a>, Shibin Wu, Hao Sun, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/cheqi\/\" target=\"_blank\" rel=\"noreferrer noopener\">Qi Chen<\/a>, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/yuqxia\/\" target=\"_blank\" rel=\"noreferrer noopener\">Yuqing Xia<\/a>, Chengmin Chi, Guoshuai Zhao, Zheng Liu, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/xingx\/\" target=\"_blank\" rel=\"noreferrer noopener\">Xing Xie<\/a>, Hao Allen Sun, Weiwei Deng, Qi Zhang, <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/maoyang\/\" target=\"_blank\" rel=\"noreferrer noopener\">Mao Yang<\/a><\/em><\/p>\n\n\n\n<p><strong>Note: this paper was named an Outstanding Paper at NeurIPS 2022<\/strong><\/p>\n\n\n\n<p>Current state-of-the-art document retrieval solutions typically follow an index-retrieve paradigm, where the index is not directly optimized towards the final target. The proposed Neural Corpus Indexer (NCI) model, instead, leverages a sequence-to-sequence architecture, which serves as a model-based index that takes a query as input and outputs the most relevant document identifiers. For the first time, we demonstrate that an end-to-end differentiable document retrieval model can significantly outperform both sparse inverted index and dense retrieval methods. Specifically, NCI achieves +17.6% and +16.8% relative enhancement for Recall@1 on NQ320k dataset and R-Precision on TriviaQA dataset respectively, and a competitive MRR without using an explicit re-ranking model. This work has received a NeurIPS 2022 Outstanding Paper award.<\/p>\n\n\n\n<p>The pipeline is composed of three stages. In the first stage, documents are encoded into semantic identifiers by the hierarchical k-means algorithm. In the second stage, a query generation model is employed to prepare <query,docids> training pairs. At the third stage, the NCI is trained with cross-entropy and consistency-based regularization losses. To further align with the hierarchical nature of the semantic identifiers, a weight adaptation mechanism is introduced to make the decoder aware of semantic prefixes. During inference, top N relevant documents can be easily obtained via beam search. The proposed approach introduces architectural and training choices that demonstrate the promising future of neural indexers as a viable alternative. And the discussed open questions can serve as an inspiration for future research.<\/p>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-4 wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button is-style-outline is-style-outline--57792e3cc7d7d665275454f89e023623\"><a data-bi-type=\"button\" class=\"wp-block-button__link\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/a-neural-corpus-indexer-for-document-retrieval\/\" target=\"_blank\" rel=\"noreferrer noopener\">Read the paper<\/a><\/div>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity is-style-dots\"\/>\n\n\n\n<h2 id=\"microsoft-research-career-opportunities-come-join-us\">Microsoft Research career opportunities \u2013 come join us!<\/h2>\n\n\n\n<p>We\u2019re hiring for multiple roles including internships and researchers at all levels in multiple Microsoft Research labs. Join us and work on causal ML, precision health, genomics, deep learning, robotics, or computational chemistry. If you\u2019re attending the conference, stop by the <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/event\/neurips-2022\/booth-schedule\/\" target=\"_blank\" rel=\"noreferrer noopener\">Microsoft booth (Expo Hall G, Booth #202)<\/a> to speak with researchers and recruiters about working at Microsoft and open job opportunities. Or you can browse our current openings at <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/event\/neurips-2022\/opportunities\/\" target=\"_blank\" rel=\"noreferrer noopener\">NeurIPS 2022 \u2013 Microsoft Research career opportunities<\/a>.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>This special edition of Research Focus highlights some of the 100+ papers from Microsoft Research that were accepted for publication at NeurIPS&nbsp;2022 \u2013 the thirty-sixth annual Conference on Neural Information Processing Systems. Dongkuan Xu, Subhabrata Mukherjee, Xiaodong Liu, Debadeepta Dey, Wenhui Wang, Xiang Zhang, Ahmed Hassan Awadallah, Jianfeng Gao Knowledge distillation (KD) is effective in [&hellip;]<\/p>\n","protected":false},"author":42183,"featured_media":901107,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"categories":[1],"tags":[],"research-area":[13556],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[243984],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-900798","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-post-option-blog-homepage-featured"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[199560],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[144916,268548,286754,470706,510017,714577],"related-projects":[],"related-events":[],"related-researchers":[],"msr_type":"Post","featured_image_thumbnail":"<img width=\"960\" height=\"540\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-960x540.jpg\" class=\"img-object-cover\" alt=\"Microsoft Research - Research Focus 05 Week of November 28th, 2022\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-1536x864.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-2048x1153.jpg 2048w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-343x193.jpg 343w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-scaled-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2022\/11\/1400x788_Research_focus5_blog_thumbnail-1920x1080.jpg 1920w\" sizes=\"(max-width: 960px) 100vw, 960px\" \/>","byline":"","formattedDate":"November 29, 2022","formattedExcerpt":"This special edition of Research Focus highlights some of the 100+ papers from Microsoft Research that were accepted for publication at NeurIPS&nbsp;2022 \u2013 the thirty-sixth annual Conference on Neural Information Processing Systems. Dongkuan Xu, Subhabrata Mukherjee, Xiaodong Liu, Debadeepta Dey, Wenhui Wang, Xiang Zhang, Ahmed&hellip;","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/900798"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/42183"}],"replies":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/comments?post=900798"}],"version-history":[{"count":46,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/900798\/revisions"}],"predecessor-version":[{"id":912105,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/900798\/revisions\/912105"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/901107"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=900798"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/categories?post=900798"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/tags?post=900798"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=900798"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=900798"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=900798"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=900798"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=900798"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=900798"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=900798"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=900798"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}