Canada

Transformer achieves promising results on various tasks. However, self-attention suffers from quadratic memory requirements with respect to the sequence length. Existing work focuses on reducing time and space complexity from an algorithm perspective. In this work, we propose sequence parallelism, a memory-efficient parallelism to solve this issue from system perspective instead. Our approach is compatible with most existing parallelisms (\eg data, pipeline, and tensor parallelism), which means our sequence parallelism makes 4D parallelism possible. More importantly, we no longer require a single device to hold the whole sequence. Besides, using efficient attention with linear complexity, our sequence parallelism enables us to train transformer with infinite long sequence. Specifically, we split the input sequence into multiple chunks and feed each chunk into its corresponding device (\ie GPU). To compute the attention output, we integrated ring-style communication with self-attention calculation and proposed Ring Self-Attention (RSA). Experiments show that sequence parallelism performs well when scaling with batch size and sequence length. Compared with tensor parallelism, our approach achieved $13.7\times$ and $3.0\times$ maximum batch size and sequence length respectively when scaling up to 64 NVIDIA P100 GPUs. With efficient attention, sequence can handle sequence with over 114K tokens, which is over $27\times$ longer than existing efficient attention works holding the whole sequence on a single device.

&lt;iframe src=&quot;https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/ad7156b0-8be8-41e1-b0a5-2d10a1e497b5&quot; width=&quot;300&quot; height=&quot;400&quot;&gt;&lt;/iframe&gt;

ACL 2023

Sequence Parallelism: Long Sequence Training from System Perspective

efficient ai system

long sequence training

transformer

**Welcome to EMNLP 2023!**

On behalf of the EMNLP 2023 Organizing Committee, I extend a warm and heartfelt welcome to all of you to the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP). It is with immense pleasure and excitement that we gather here in Singapore, a vibrant hub of innovation and technological advancement.

The conference program is packed with insightful presentations, thought-provoking workshops, and engaging networking opportunities. In addition to the technical sessions, we have also planned several social events that will provide you with the opportunity to connect with your colleagues.

The conference is held in person at the Resorts World Convention Centre in Singapore, and available on-line with the help of Underline.

For those in Singapore, I hope that you will find time to explore the exciting Sentosa Island, Gardens by the Bay, Singapore Botanic Gardens and the many other attractions unique to Singapore.

Of course, we are grateful to our sponsors and partners for their generous support of EMNLP 2023, whose contributions make it possible for us to host this world-class event.

Yuji Matsumoto (RIKEN AIP)<br>
EMNLP 2023 General Chair<br><br><br><br>

To access the **EMNLP 2023** event page on Underline, you need to register for the Conference.<br>
Please follow **[this link](https://2023.emnlp.org/registration/)** for more details. 

EMNLP 2023

EMNLP 2023 took place in Singapore from Dec 6th to Dec 10th, 2023.

Transformer achieves promising results on various tasks. However, self-attention suffers from quadratic memory requirements with respect to the sequence length. Existing work focuses on reducing time and space complexity from an algorithm perspective. In this work, we propose sequence parallelism, a memory-efficient parallelism to solve this issue from system perspective instead. Our approach is compatible with most existing parallelisms (\eg data, pipeline, and tensor parallelism), which means our sequence parallelism makes 4D parallelism possible. More importantly, we no longer require a single device to hold the whole sequence. Besides, using efficient attention with linear complexity, our sequence parallelism enables us to train transformer with infinite long sequence. Specifically, we split the input sequence into multiple chunks and feed each chunk into its corresponding device (\ie GPU). To compute the attention output, we integrated ring-style communication with self-attention calculation and proposed Ring Self-Attention (RSA). Experiments show that sequence parallelism performs well when scaling with batch size and sequence length. Compared with tensor parallelism, our approach achieved $13.7\times$ and $3.0\times$ maximum batch size and sequence length respectively when scaling up to 64 NVIDIA P100 GPUs. With efficient attention, sequence can handle sequence with over 114K tokens, which is over $27\times$ longer than existing efficient attention works holding the whole sequence on a single device.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/ad7156b0-8be8-41e1-b0a5-2d10a1e497b5" width="300" height="400"></iframe>

poster

### Welcome to ACL 2023, the 61st Annual Meeting of the Association for Computational Linguistics! 
<br>The conference will be held in Toronto, Canada, July 9-14, 2023.<br><br>
Following the succession of the recent conferences in our field, ACL 2023 will adopt a hybrid format.
While the impact of Covid has considerably diminished in terms of traveling, obtaining visas to Canada
entails a very long process. Moreover, the global economic conditions pose challenges for many individuals to travel to conferences. Recognizing these circumstances, we know many participants may not be
able to attend the conference in person. Therefore, we are committed to providing a great virtual platform
so everyone has the opportunity to interact with other participants and enjoy the conference. Based on the
current registered participants, approxiately 30% have chosen to attend the conference virtually. Whether
you join us in person or virtually, we sincerely hope everyone has a remarkable conference experience.<br>
This General Chair’s message is where I express my gratitude to the many individuals who have made
enormous contributions to the conference over the past year.

Read [**ACL 2023 General Chair's message**](https://docs.google.com/document/d/1WobYM7norbG4dI48s75HfJoD89qgX5a_F-6U8AteLSA/edit?usp=sharing/) in full.

##### **[Conference Handbook](https://2023.aclweb.org/downloads/acl2023-handbook.pdf)**
<br><br><br><br>

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing.

The anonymity on the Darknet allows vendors to stay undetected by using multiple vendor aliases or frequently migrating between markets. Consequently, illegal markets and their connections are challenging to uncover on the Darknet. To identify relationships between illegal markets and their vendors, we propose VendorLink, an NLP-based approach that examines writing patterns to verify, identify, and link unique vendor accounts across text advertisements (ads) on seven public Darknet markets. In contrast to existing literature, VendorLink utilizes the strength of supervised pre-training to perform closed-set vendor verification, open-set vendor identification, and low-resource market adaption tasks. Through VendorLink, we uncover (i) 15 migrants and 71 potential aliases in the Alphabay-Dreams-Silk dataset, (ii) 17 migrants and 3 potential aliases in the Valhalla-Berlusconi dataset, and (iii) 75 migrants and 10 potential aliases in the Traderoute-Agora dataset. Altogether, our approach can help Law Enforcement Agencies (LEA) make more informed decisions by verifying and identifying migrating vendors and their potential aliases on existing and Low-Resource (LR) emerging Darknet markets.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/664f4bd4-33bb-44fc-ac41-8edeb77e978d" width="300" height="400"></iframe>

VendorLink: An NLP approach for Identifying & Linking Vendor Migrants & Potential Aliases on Darknet Markets

Prior work on language models (LMs) shows that training on a large number of diverse tasks improves few-shot learning (FSL) performance on new tasks. We take this to the extreme, automatically extracting 413,299 tasks from internet tables - orders of magnitude more than the next-largest public datasets. Finetuning on the resulting dataset leads to improved FSL performance on Natural Language Processing (NLP) tasks, but not proportionally to dataset scale. In fact, we find that narrow subsets of our dataset sometimes outperform more diverse datasets. For example, finetuning on software documentation from support.google.com raises FSL performance by a mean of +7.5% on 52 downstream tasks, which beats training on 40 human-curated NLP datasets (+6.7%). Finetuning on various narrow datasets leads to similar broad improvements across test tasks, suggesting that the gains are not from domain adaptation but adapting to FSL in general. We do not observe clear patterns between the datasets that lead to FSL gains, leaving open questions about why certain data helps with FSL.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/bd2750b1-b554-4934-8b00-23dfbc084fc5" width="300" height="400"></iframe>

Few-shot Adaptation Works with UnpredicTable Data

Recently, question answering over temporal knowledge graphs (i.e., TKGQA) has been introduced and investigated, in quest of reasoning about dynamic factual knowledge. To foster research on TKGQA, a few datasets have been curated (e.g., CronQuestions and Complex-CronQuestions), and various models have been proposed based on these datasets. Nevertheless, existing efforts overlook the fact that real-life applications of TKGQA also tend to be complex in temporal granularity, i.e., the questions may concern mixed temporal granularities (e.g., both day and month). To overcome the limitation, in this paper, we motivate the notion of multi-granularity temporal question answering over knowledge graphs and present a large scale dataset for multi-granularity TKGQA, namely MultiTQ. To the best of our knowledge, MultiTQis among the first of its kind, and compared with existing datasets on TKGQA, MultiTQfeatures at least two desirable aspectsample relevant facts and multiple temporal granularities. It is expected to better reflect real-world challenges, and serve as a test bed for TKGQA models. In addition, we propose a competing baseline MultiQA over MultiTQ, which is experimentally demonstrated to be effective in dealing with TKGQA. The data and code are released at https://github.com/czy1999/MultiTQ.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/389d7785-b0e5-4db4-a538-7064f9afb729" width="300" height="400"></iframe>

Multi-granularity Temporal Question Answering over Knowledge Graphs

The longstanding goal of multi-lingual learning has been to develop a universal cross-lingual model that can withstand the changes in multi-lingual data distributions. There has been a large amount of work to adapt such multi-lingual models to unseen target languages. However, the majority of work in this direction focuses on the standard one-hop transfer learning pipeline from source to target languages, whereas in realistic scenarios, new languages can be incorporated at any time in a sequential manner. In this paper, we present a principled Cross-lingual Continual Learning (CCL) evaluation paradigm, where we analyze different categories of approaches used to continually adapt to emerging data from different languages. We provide insights into what makes multilingual sequential learning particularly challenging. To surmount such challenges, we benchmark a representative set of cross-lingual continual learning algorithms and analyze their knowledge preservation, accumulation, and generalization capabilities compared to baselines on carefully curated datastreams. The implications of this analysis include a recipe for how to measure and balance different cross-lingual continual learning desiderata, which go beyond conventional transfer learning.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/66d2752e-338e-4bd7-893b-c3f0c0a78273" width="300" height="400"></iframe>

Cross-lingual Continual Learning

Moral norms vary across cultures. A recent line of work suggests that English large language models contain human-like moral biases, but these studies typically do not examine moral variation in a diverse cultural setting. We investigate the extent to which monolingual English language models contain knowledge about moral norms in different countries. We consider two levels of analysis: 1) whether language models capture fine-grained moral variation across countries over a variety of topics such as "homosexuality" and "divorce"; 2) whether language models capture cultural diversity and shared tendencies in which topics people around the globe tend to diverge or agree on in their moral judgment. We perform our analyses with two public datasets from the World Values Survey (across 55 countries) and PEW global surveys (across 40 countries) on morality. We find that pre-trained English language models predict empirical moral norms across countries worse than the English moral norms reported previously. However, fine-tuning language models on the survey data improves inference across countries at the expense of a less accurate estimate of the English moral norms. We discuss the relevance and challenges of incorporating cultural knowledge into the automated inference of moral norms.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/4ec3a9df-fdf2-4455-85ed-55f8e2621f83" width="300" height="400"></iframe>

Knowledge of cultural moral norms in large language models

Dataset distillation aims to create a small dataset of informative synthetic samples to rapidly train neural networks that retain the performance of the original dataset. In this paper, we focus on constructing distilled few-shot datasets for natural language processing (NLP) tasks to fine-tune pre-trained transformers. Specifically, we propose to introduce attention labels, which can efficiently distill the knowledge from the original dataset and transfer it to the transformer models via attention probabilities. We evaluated our dataset distillation methods in four various NLP tasks and demonstrated that it is possible to create distilled few-shot datasets with the attention labels, yielding impressive performances for fine-tuning BERT. Specifically, in AGNews, a four-class news classification task, our distilled few-shot dataset achieved up to 93.2% accuracy, which is 98.5% performance of the original dataset even with only one sample per class and only one gradient step.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/a6d5257b-7e67-4b06-ac8d-93ccc90efd3d" width="300" height="400"></iframe>

Dataset Distillation with Attention Labels for Fine-tuning BERT

Standard language model training employs gold human documents or human-human interaction data, and treats all training data as positive examples. Growing evidence shows that even with very large amounts of positive training data, issues remain that can be alleviated with relatively small amounts of negative data -- examples of what the model should not do. In this work, we propose a novel procedure to train with such data called the "CRINGE" loss (ContRastive Iterative Negative GEneration). We show the effectiveness of this approach across three different experiments on the tasks of safe generation, contradiction avoidance, and open-domain dialogue. Our models outperform multiple strong baselines and are conceptually simple, easy to train and implement.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/24e1a310-00e4-4b71-b5ed-37ff73408a0b" width="300" height="400"></iframe>

The CRINGE Loss: Learning what language not to model

Moral rhetoric influences our judgement. Although social scientists recognize moral expression as domain specific, there are no systematic methods for analyzing whether a text classifier learns the domain-specific expression of moral language or not. We propose Tomea, a method to compare a supervised classifier's representation of moral rhetoric across domains. Tomea enables quantitative and qualitative comparisons of moral rhetoric via an interpretable exploration of similarities and differences across moral concepts and domains. We apply Tomea on moral narratives in thirty-five thousand tweets from seven domains. We extensively evaluate the method via a crowd study, a series of cross-domain moral classification comparisons, and a qualitative analysis of cross-domain moral expression.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/f4f925a4-6568-4e9e-b061-84671aae32ee" width="300" height="400"></iframe>

What does a Text Classifier Learn about Morality? An Explainable Method for Cross-Domain Comparison of Moral Rhetoric

Natural language inference (NLI) models are susceptible to learning shortcuts, i.e. decision rules that spuriously correlate with the label. As a result, they achieve high in-distribution performance, but fail to generalize to out-of-distribution samples where such correlations do not hold. In this paper, we present a training method to reduce the reliance of NLI models on shortcuts and improve their out-of-distribution performance without assuming prior knowledge of the shortcuts being targeted. To this end, we propose a minimax objective between a learner model being trained for the NLI task, and an auxiliary model aiming to maximize the learner's loss by up-weighting examples from regions of the input space where the learner incurs high losses. This process incentivizes the learner to focus on under-represented "hard" examples with patterns that contradict the shortcuts learned from the prevailing "easy" examples. Experimental results on three NLI datasets demonstrate that our method consistently outperforms other robustness enhancing techniques on out-of-distribution adversarial test sets, while maintaining high in-distribution accuracy.

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/c18c06ec-bb6d-4090-b1a8-2d4686d389d3" width="300" height="400"></iframe>

Improving the robustness of NLI models with minimax training

In this paper, we propose SimLM (Similarity matching with Language Model pre-training), a simple yet effective pre-training method for dense passage retrieval. It employs a simple bottleneck architecture that learns to compress the passage information into a dense vector through self-supervised pre-training. We use a replaced language modeling objective, which is inspired by ELECTRA (Clark et al., 2020), to improve the sample efficiency and reduce the mismatch of the input distribution between pre-training and fine-tuning. SimLM only requires access to an unlabeled corpus and is more broadly applicable when there are no labeled data or queries. We conduct experiments on several large-scale passage retrieval datasets and show substantial improvements over strong baselines under various settings. Remarkably, SimLM even outperforms multi-vector approaches such as ColBERTv2 (Santhanam et al., 2021) which incurs significantly more storage cost. Our code and model checkpoints are available at https://github.com/microsoft/unilm/tree/master/simlm .

<iframe src="https://app.sli.do/event/1kgCXjyZMgdV89PsksaJsj/embed/polls/560e7df0-8c79-4097-b996-9dea86bf5189" width="300" height="400"></iframe>

Downloads

Next from ACL 2023

VendorLink: An NLP approach for Identifying & Linking Vendor Migrants & Potential Aliases on Darknet Markets

Similar lecture

Introducing Rhetorical Parallelism Detection: A New Task with Datasets, Metrics, and Baselines