Eduardo Sánchez, Belen Alastruey, Christophe Ropers, Pontus Stenetorp, Mikel Artetxe, Marta R. Costa-jussà
arXiv preprint
We propose a new benchmark to measure a language model's linguistic reasoning skills without relying on pre-existing language-specific knowledge. The test covers 894 questions grouped in 160 problems across 75 (mostly) extremely low-resource languages, extracted from the International Linguistic Olympiad corpus. To attain high accuracy on this benchmark, models don't need previous knowledge of the tested language, as all the information needed to solve the linguistic puzzle is presented in the context. We find that, while all analyzed models rank below 25% accuracy, there is a significant gap between open and closed models, with the best-performing proprietary model at 24.05% and the best-performing open model at 8.84%.
@misc{sanchez2024linguini
title = {Linguini: A benchmark for language-agnostic linguistic reasoning},
author = {Sánchez, Eduardo and Alastruey, Belen and Ropers, Christophe and Stenetorp, Pontus and Artetxe, Mikel and Costa-jussà, Marta R.},
year = {2024},
month = {09},
publisher = {arXiv},
url = {https://arxiv.org/abs/2409.12126},
}
Julen Etxaniz, Oscar Sainz, Naiara Miguel, Itziar Aldabe, German Rigau, Eneko Agirre, Aitor Ormazabal, Mikel Artetxe, Aitor Soroa
ACL 2024
We introduce Latxa, a family of large language models for Basque ranging from 7 to 70 billion parameters. Latxa is based on Llama 2, which we continue pretraining on a new Basque corpus comprising 4.3M documents and 4.2B tokens. Addressing the scarcity of high-quality benchmarks for Basque, we further introduce 4 multiple choice evaluation datasets: EusProficiency, comprising 5,169 questions from official language proficiency exams; EusReading, comprising 352 reading comprehension questions; EusTrivia, comprising 1,715 trivia questions from 5 knowledge areas; and EusExams, comprising 16,046 questions from public examinations. In our extensive evaluation, Latxa outperforms all previous open models we compare to by a large margin. In addition, it is competitive with GPT-4 Turbo in language proficiency and understanding, despite lagging behind in reading comprehension and knowledge-intensive tasks. Both the Latxa family of models, as well as our new pretraining corpora and evaluation datasets, are publicly available under open licenses. Our suite enables reproducible research on methods to build LLMs for low-resource languages.
@inproceedings{etxaniz2024latxa
title = {Latxa: An Open Language Model and Evaluation Suite for Basque},
author = {Etxaniz, Julen and Sainz, Oscar and Miguel, Naiara and Aldabe, Itziar and Rigau, German and Agirre, Eneko and Ormazabal, Aitor and Artetxe, Mikel and Soroa, Aitor},
booktitle = {Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
pages = {14952-14972},
year = {2024},
month = {08},
address = {Bangkok, Thailand},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2024.acl-long.799},
url = {https://aclanthology.org/2024.acl-long.799},
}
Lucas Bandarkar, Davis Liang, Benjamin Muller, Mikel Artetxe, Satya Narayan Shukla, Donald Husa, Naman Goyal, Abhinandan Krishnan, Luke Zettlemoyer, Madian Khabsa
ACL 2024
We present Belebele, a multiple-choice machine reading comprehension (MRC) dataset spanning 122 language variants. Significantly expanding the language coverage of natural language understanding (NLU) benchmarks, this dataset enables the evaluation of text models in high-, medium-, and low-resource languages. Each question is based on a short passage from the FLORES-200 dataset and has four multiple-choice answers. The questions were carefully curated to discriminate between models with different levels of general language comprehension. The English dataset on its own proves difficult enough to challenge state-of-the-art language models. Being fully parallel, this dataset enables direct comparison of model performance across all languages. We use this dataset to evaluate the capabilities of multilingual masked language models (MLMs) and large language models (LLMs). We present extensive results and findings, notably that despite significant cross-lingual transfer in English-centric LLMs, much smaller MLMs pretrained on balanced multilingual data still understand far more languages. Overall, Belebele opens up new avenues for evaluating and analyzing the multilingual capabilities of NLP systems.
@inproceedings{bandarkar2024belebele
title = {The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants},
author = {Bandarkar, Lucas and Liang, Davis and Muller, Benjamin and Artetxe, Mikel and Shukla, Satya Narayan and Husa, Donald and Goyal, Naman and Krishnan, Abhinandan and Zettlemoyer, Luke and Khabsa, Madian},
booktitle = {Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
pages = {749-775},
year = {2024},
month = {08},
address = {Bangkok, Thailand},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2024.acl-long.44},
url = {https://aclanthology.org/2024.acl-long.44},
}
Julen Etxaniz, Gorka Azkune, Aitor Soroa, Oier Lacalle, Mikel Artetxe
NAACL 2024
Translate-test is a popular technique to improve the performance of multilingual language models. This approach works by translating the input into English using an external machine translation system before running inference. However, these improvements can be attributed to the use of a separate translation system, which is typically trained on large amounts of parallel data not seen by the language model. In this work, we introduce a new approach called self-translate that leverages the few-shot translation capabilities of multilingual language models. This allows us to analyze the effect of translation in isolation. Experiments over 5 tasks show that self-translate consistently outperforms direct inference, demonstrating that language models are unable to leverage their full multilingual potential when prompted in non-English languages. Our code is available at https://github.com/juletx/self-translate.
@inproceedings{etxaniz2024multilingual
title = {Do Multilingual Language Models Think Better in English?},
author = {Etxaniz, Julen and Azkune, Gorka and Soroa, Aitor and Lacalle, Oier and Artetxe, Mikel},
booktitle = {Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers)},
pages = {550-564},
year = {2024},
month = {06},
address = {Mexico City, Mexico},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2024.naacl-short.46},
url = {https://aclanthology.org/2024.naacl-short.46},
}
Ahmed Elhady, Khaled Elsayed, Eneko Agirre, Mikel Artetxe
NAACL 2024
Factual accuracy is an important property of neural abstractive summarization models, especially in fact-critical domains such as the clinical literature. In this work, we introduce a guided continued pre-training stage for encoder-decoder models that improves their understanding of the factual attributes of documents, which is followed by supervised fine-tuning on summarization. Our approach extends the pre-training recipe of BART to incorporate 3 additional objectives based on PICO spans, which capture the population, intervention, comparison, and outcomes related to a clinical study. Experiments on multi-document summarization in the clinical domain demonstrate that our approach is competitive with prior work, improving the quality and factuality of the summaries and achieving the best-published results in factual accuracy on the MSLR task.
@inproceedings{elhady2024improving
title = {Improving Factuality in Clinical Abstractive Multi-Document Summarization by Guided Continued Pre-training},
author = {Elhady, Ahmed and Elsayed, Khaled and Agirre, Eneko and Artetxe, Mikel},
booktitle = {Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers)},
pages = {755-761},
year = {2024},
month = {06},
address = {Mexico City, Mexico},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2024.naacl-short.66},
url = {https://aclanthology.org/2024.naacl-short.66},
}
Julen Etxaniz, Gorka Azkune, Aitor Soroa, Oier Lopez de Lacalle, Mikel Artetxe
arXiv preprint
Large Language Models (LLMs) exhibit extensive knowledge about the world, but most evaluations have been limited to global or anglocentric subjects. This raises the question of how well these models perform on topics relevant to other cultures, whose presence on the web is not that prominent. To address this gap, we introduce BertaQA, a multiple-choice trivia dataset that is parallel in English and Basque. The dataset consists of a local subset with questions pertinent to the Basque culture, and a global subset with questions of broader interest. We find that state-of-the-art LLMs struggle with local cultural knowledge, even as they excel on global topics. However, we show that continued pre-training in Basque significantly improves the models' performance on Basque culture, even when queried in English. To our knowledge, this is the first solid evidence of knowledge transfer from a low-resource to a high-resource language. Our analysis sheds light on the complex interplay between language and knowledge, and reveals that some prior findings do not fully hold when reassessed on local topics. Our dataset and evaluation code are available under open licenses at https://github.com/juletx/BertaQA.
@misc{etxaniz2024bertaqa
title = {BertaQA: How Much Do Language Models Know About Local Culture?},
author = {Etxaniz, Julen and Azkune, Gorka and Soroa, Aitor and de Lacalle, Oier Lopez and Artetxe, Mikel},
year = {2024},
month = {06},
publisher = {arXiv},
url = {https://arxiv.org/abs/2406.07302},
}
Piotr Padlewski, Max Bain, Matthew Henderson, Zhongkai Zhu, Nishant Relan, Hai Pham, Donovan Ong, Kaloyan Aleksiev, Aitor Ormazabal, Samuel Phua, Ethan Yeo, Eugenie Lamprecht, Qi Liu, Yuqi Wang, Eric Chen, Deyu Fu, Lei Li, Che Zheng, Cyprien de Masson d'Autume, Dani Yogatama, Mikel Artetxe, Yi Tay
arXiv preprint
We introduce Vibe-Eval: a new open benchmark and framework for evaluating multimodal chat models. Vibe-Eval consists of 269 visual understanding prompts, including 100 of hard difficulty, complete with gold-standard responses authored by experts. Vibe-Eval is open-ended and challenging with dual objectives: (i) vibe checking multimodal chat models for day-to-day tasks and (ii) rigorously testing and probing the capabilities of present frontier models. Notably, our hard set contains >50% questions that all frontier models answer incorrectly. We explore the nuances of designing, evaluating, and ranking models on ultra challenging prompts. We also discuss trade-offs between human and automatic evaluation, and show that automatic model evaluation using Reka Core roughly correlates to human judgment. We offer free API access for the purpose of lightweight evaluation and plan to conduct formal human evaluations for public models that perform well on the Vibe-Eval's automatic scores. We release the evaluation code and data, see https://github.com/reka-ai/reka-vibe-eval
@misc{padlewski2024vibeeval
title = {Vibe-Eval: A hard evaluation suite for measuring progress of multimodal language models},
author = {Padlewski, Piotr and Bain, Max and Henderson, Matthew and Zhu, Zhongkai and Relan, Nishant and Pham, Hai and Ong, Donovan and Aleksiev, Kaloyan and Ormazabal, Aitor and Phua, Samuel and Yeo, Ethan and Lamprecht, Eugenie and Liu, Qi and Wang, Yuqi and Chen, Eric and Fu, Deyu and Li, Lei and Zheng, Che and de Masson d'Autume, Cyprien and Yogatama, Dani and Artetxe, Mikel and Tay, Yi},
year = {2024},
month = {05},
publisher = {arXiv},
url = {https://arxiv.org/abs/2405.02287},
}
Reka Team, Aitor Ormazabal, Che Zheng, Cyprien de Masson d'Autume, Dani Yogatama, Deyu Fu, Donovan Ong, Eric Chen, Eugenie Lamprecht, Hai Pham, Isaac Ong, Kaloyan Aleksiev, Lei Li, Matthew Henderson, Max Bain, Mikel Artetxe, Nishant Relan, Piotr Padlewski, Qi Liu, Ren Chen, Samuel Phua, Yazheng Yang, Yi Tay, Yuqi Wang, Zhongkai Zhu, Zhihui Xie
arXiv preprint
We introduce Reka Core, Flash, and Edge, a series of powerful multimodal language models trained from scratch by Reka. Reka models are able to process and reason with text, images, video, and audio inputs. This technical report discusses details of training some of these models and provides comprehensive evaluation results. We show that Reka Edge and Reka Flash are not only state-of-the-art but also outperform many much larger models, delivering outsized values for their respective compute class. Meanwhile, our most capable and largest model, Reka Core, approaches the best frontier models on both automatic evaluations and blind human evaluations. On image question answering benchmarks (e.g. MMMU, VQAv2), Core performs competitively to GPT4-V. Meanwhile, on multimodal chat, Core ranks as the second most preferred model under a blind third-party human evaluation setup, outperforming other models such as Claude 3 Opus. On text benchmarks, Core not only performs competitively to other frontier models on a set of well-established benchmarks (e.g. MMLU, GSM8K) but also outperforms GPT4-0613 on human evaluation. On video question answering (Perception-Test), Core outperforms Gemini Ultra. Models are shipped in production at http://chat.reka.ai . A showcase of non cherry picked qualitative examples can also be found at http://showcase.reka.ai .
@misc{rekateam2024reka
title = {Reka Core, Flash, and Edge: A Series of Powerful Multimodal Language Models},
author = {Reka Team and Ormazabal, Aitor and Zheng, Che and de Masson d'Autume, Cyprien and Yogatama, Dani and Fu, Deyu and Ong, Donovan and Chen, Eric and Lamprecht, Eugenie and Pham, Hai and Ong, Isaac and Aleksiev, Kaloyan and Li, Lei and Henderson, Matthew and Bain, Max and Artetxe, Mikel and Relan, Nishant and Padlewski, Piotr and Liu, Qi and Chen, Ren and Phua, Samuel and Yang, Yazheng and Tay, Yi and Wang, Yuqi and Zhu, Zhongkai and Xie, Zhihui},
year = {2024},
month = {04},
publisher = {arXiv},
url = {https://arxiv.org/abs/2404.12387},
}
Yihong Chen, Kelly Marchisio, Roberta Raileanu, David Ifeoluwa Adelani, Pontus Stenetorp, Sebastian Riedel, Mikel Artetxe
NeurIPS 2023
Pretrained language models (PLMs) are today the primary model for natural language processing. Despite their impressive downstream performance, it can be difficult to apply PLMs to new languages, a barrier to making their capabilities universally accessible. While prior work has shown it possible to address this issue by learning a new embedding layer for the new language, doing so is both data and compute inefficient. We propose to use an active forgetting mechanism during pretraining, as a simple way of creating PLMs that can quickly adapt to new languages. Concretely, by resetting the embedding layer every K updates during pretraining, we encourage the PLM to improve its ability of learning new embeddings within limited number of updates, similar to a meta-learning effect. Experiments with RoBERTa show that models pretrained with our forgetting mechanism not only demonstrate faster convergence during language adaptation, but also outperform standard ones in a low-data regime, particularly for languages that are distant from English. Code will be available at https://github.com/facebookresearch/language-model-plasticity.
@inproceedings{chen2023improving
title = {Improving Language Plasticity via Pretraining with Active Forgetting},
author = {Chen, Yihong and Marchisio, Kelly and Raileanu, Roberta and Adelani, David Ifeoluwa and Stenetorp, Pontus and Riedel, Sebastian and Artetxe, Mikel},
booktitle = {Proceedings of the 37th Conference on Neural Information Processing Systems (NeurIPS 2023)},
pages = {31543-31557},
year = {2023},
month = {12},
publisher = {Curran Associates, Inc.},
url = {https://papers.nips.cc/paper_files/paper/2023/hash/6450ea28ebbc8437bc38775157818172-Abstract-Conference.html},
}
Mikel Artetxe, Vedanuj Goswami, Shruti Bhosale, Angela Fan, Luke Zettlemoyer
EMNLP 2023
Machine Translation (MT) has been widely used for cross-lingual classification, either by translating the test set into English and running inference with a monolingual model (translate-test), or translating the training set into the target languages and finetuning a multilingual model (translate-train). However, most research in the area focuses on the multilingual models rather than the MT component. We show that, by using a stronger MT system and mitigating the mismatch between training on original text and running inference on machine translated text, translate-test can do substantially better than previously assumed. The optimal approach, however, is highly task dependent, as we identify various sources of cross-lingual transfer gap that affect different tasks and approaches differently. Our work calls into question the dominance of multilingual models for cross-lingual classification, and prompts to pay more attention to MT-based baselines.
@inproceedings{artetxe2023revisiting
title = {Revisiting Machine Translation for Cross-lingual Classification},
author = {Artetxe, Mikel and Goswami, Vedanuj and Bhosale, Shruti and Fan, Angela and Zettlemoyer, Luke},
booktitle = {Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing},
pages = {6489-6499},
year = {2023},
month = {12},
address = {Singapore},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2023.emnlp-main.399},
url = {https://aclanthology.org/2023.emnlp-main.399},
}
Aitor Ormazabal, Mikel Artetxe, Eneko Agirre
EMNLP 2023
Methods for adapting language models (LMs) to new tasks and domains have traditionally assumed white-box access to the model, and work by modifying its parameters. However, this is incompatible with a recent trend in the field, where the highest quality models are only available as black-boxes through inference APIs. Even when the model weights are available, the computational cost of fine-tuning large LMs can be prohibitive for most practitioners. In this work, we present a lightweight method for adapting large LMs to new domains and tasks, assuming no access to their weights or intermediate activations. Our approach fine-tunes a small white-box LM and combines it with the large black-box LM at the probability level through a small network, learned on a small validation set. We validate our approach by adapting a large LM (OPT-30B) to several domains and a downstream task (machine translation), observing improved performance in all cases, of up to 9%, while using a domain expert 23x smaller.
@inproceedings{ormazabal2023comblm
title = {CombLM: Adapting Black-Box Language Models through Small Fine-Tuned Models},
author = {Ormazabal, Aitor and Artetxe, Mikel and Agirre, Eneko},
booktitle = {Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing},
pages = {2961-2974},
year = {2023},
month = {12},
address = {Singapore},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2023.emnlp-main.180},
url = {https://aclanthology.org/2023.emnlp-main.180},
}
Dieuwke Hupkes, Mario Giulianelli, Verna Dankers, Mikel Artetxe, Yanai Elazar, Tiago Pimentel, Christos Christodoulopoulos, Karim Lasri, Naomi Saphra, Arabella Sinclair, Dennis Ulmer, Florian Schottmann, Khuyagbaatar Batsuren, Kaiser Sun, Koustuv Sinha, Leila Khalatbari, Maria Ryskina, Rita Frieske, Ryan Cotterell, Zhijing Jin
Nature Machine Intelligence
The ability to generalize well is one of the primary desiderata for models of natural language processing (NLP), but what ‘good generalization’ entails and how it should be evaluated is not well understood. In this Analysis we present a taxonomy for characterizing and understanding generalization research in NLP. The proposed taxonomy is based on an extensive literature review and contains five axes along which generalization studies can differ: their main motivation, the type of generalization they aim to solve, the type of data shift they consider, the source by which this data shift originated, and the locus of the shift within the NLP modelling pipeline. We use our taxonomy to classify over 700 experiments, and we use the results to present an in-depth analysis that maps out the current state of generalization research in NLP and make recommendations for which areas deserve attention in the future.
@article{hupkes2023taxonomy
title = {A taxonomy and review of generalization research in NLP},
author = {Hupkes, Dieuwke and Giulianelli, Mario and Dankers, Verna and Artetxe, Mikel and Elazar, Yanai and Pimentel, Tiago and Christodoulopoulos, Christos and Lasri, Karim and Saphra, Naomi and Sinclair, Arabella and Ulmer, Dennis and Schottmann, Florian and Batsuren, Khuyagbaatar and Sun, Kaiser and Sinha, Koustuv and Khalatbari, Leila and Ryskina, Maria and Frieske, Rita and Cotterell, Ryan and Jin, Zhijing},
journal = {Nature Machine Intelligence},
volume = {5},
number = {10},
pages = {1161-1174},
year = {2023},
month = {10},
issn = {2522-5839},
doi = {10.1038/s42256-023-00729-y},
url = {https://doi.org/10.1038/s42256-023-00729-y},
}
Eduardo Sánchez, Pierre Andrews, Pontus Stenetorp, Mikel Artetxe, Marta R. Costa-jussà
arXiv preprint
While machine translation (MT) systems have seen significant improvements, it is still common for translations to reflect societal biases, such as gender bias. Decoder-only Large Language Models (LLMs) have demonstrated potential in MT, albeit with performance slightly lagging behind traditional encoder-decoder Neural Machine Translation (NMT) systems. However, LLMs offer a unique advantage: the ability to control the properties of the output through prompts. In this study, we leverage this flexibility to explore LLaMa's capability to produce gender-specific translations. Our results indicate that LLaMa can generate gender-specific translations with translation accuracy and gender bias comparable to NLLB, a state-of-the-art multilingual NMT system. Furthermore, our experiments reveal that LLaMa's gender-specific translations rely on coreference resolution to determine gender, showing higher gender variance in gender-ambiguous datasets but maintaining consistency in less ambiguous contexts. This research investigates the potential and challenges of using LLMs for gender-specific translations as an instance of the controllability of outputs offered by LLMs.
@misc{sanchez2023genderspecific
title = {Gender-specific Machine Translation with Large Language Models},
author = {Sánchez, Eduardo and Andrews, Pierre and Stenetorp, Pontus and Artetxe, Mikel and Costa-jussà, Marta R.},
year = {2023},
month = {09},
publisher = {arXiv},
url = {https://arxiv.org/abs/2309.03175},
}
Anirudh Mittal, Timo Schick, Mikel Artetxe, Jane Dwivedi-Yu
arXiv preprint
As increasingly sophisticated language models emerge, their trustworthiness becomes a pivotal issue, especially in tasks such as summarization and question-answering. Ensuring their responses are contextually grounded and faithful is challenging due to the linguistic diversity and the myriad of possible answers. In this paper, we introduce a novel approach to evaluate faithfulness of machine-generated text by computing the longest noncontinuous substring of the claim that is supported by the context, which we refer to as the Longest Supported Subsequence (LSS). Using a new human-annotated dataset, we finetune a model to generate LSS. We introduce a new method of evaluation and demonstrate that these metrics correlate better with human ratings when LSS is employed, as opposed to when it is not. Our proposed metric demonstrates an 18% enhancement over the prevailing state-of-the-art metric for faithfulness on our dataset. Our metric consistently outperforms other metrics on a summarization dataset across six different models. Finally, we compare several popular Large Language Models (LLMs) for faithfulness using this metric. We release the human-annotated dataset built for predicting LSS and our fine-tuned model for evaluating faithfulness.
@misc{mittal2023evaluation
title = {Evaluation of Faithfulness Using the Longest Supported Subsequence},
author = {Mittal, Anirudh and Schick, Timo and Artetxe, Mikel and Dwivedi-Yu, Jane},
year = {2023},
month = {08},
publisher = {arXiv},
url = {https://arxiv.org/abs/2308.12157},
}
Mengzhou Xia, Mikel Artetxe, Chunting Zhou, Xi Victoria Lin, Ramakanth Pasunuru, Danqi Chen, Luke Zettlemoyer, Veselin Stoyanov
ACL 2023
Scaling up language models has led to unprecedented performance gains, but little is understood about how the training dynamics change as models get larger. How do language models of different sizes learn during pre-training? Why do larger language models demonstrate more desirable behaviors? In this paper, we analyze the intermediate training checkpoints of differently sized OPT models (Zhang et al., 2022)—from 125M to 175B parameters—on next-token prediction, sequence-level generation and downstream tasks. We find that 1) at a given perplexity and independent of model sizes, a similar subset of training tokens see the most significant reduction in loss, with the rest stagnating or showing double-descent behavior (Nakkiran et al., 2020); 2) early in training, all models learn to reduce the perplexity of grammatical sequences that contain hallucinations, with small models halting at this suboptimal distribution and larger ones eventually learning to assign these sequences lower probabilities; and 3) perplexity is a strong predictor of in-context learning performance on 74 multiple-choice tasks from BIG-Bench, and this holds independent of the model size. Together, these results show that perplexity is more predictive of model behaviors than model size or training computation.
@inproceedings{xia2023training
title = {Training Trajectories of Language Models Across Scales},
author = {Xia, Mengzhou and Artetxe, Mikel and Zhou, Chunting and Lin, Xi Victoria and Pasunuru, Ramakanth and Chen, Danqi and Zettlemoyer, Luke and Stoyanov, Veselin},
booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
pages = {13711-13738},
year = {2023},
month = {07},
address = {Toronto, Canada},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2023.acl-long.767},
url = {https://aclanthology.org/2023.acl-long.767},
}
Kelly Marchisio, Patrick Lewis, Yihong Chen, Mikel Artetxe
Findings of ACL 2023
Prior work shows that it is possible to expand pretrained Masked Language Models (MLMs) to new languages by learning a new set of embeddings, while keeping the transformer body frozen. Despite learning a small subset of parameters, this approach is not compute-efficient, as training the new embeddings requires a full forward and backward pass over the entire model. We propose mini-model adaptation, a compute-efficient alternative that builds a shallow mini-model from a fraction of a large model’s parameters. New language-specific embeddings can then be efficiently trained over the mini-model and plugged into the aligned large model for rapid cross-lingual transfer. We explore two approaches to learn mini-models: MINIJOINT, which jointly pretrains the primary model and the mini-model using a single transformer with a secondary MLM head at a middle layer; and MINIPOST, where we start from a regular pretrained model, build a mini-model by extracting and freezing a few layers, and learn a small number of parameters on top. Experiments on XNLI, MLQA and PAWS-X show that mini-model adaptation matches the performance of the standard approach using up to 2.3x less compute on average.
@inproceedings{marchisio2023minimodel
title = {Mini-Model Adaptation: Efficiently Extending Pretrained Models to New Languages via Aligned Shallow Training},
author = {Marchisio, Kelly and Lewis, Patrick and Chen, Yihong and Artetxe, Mikel},
booktitle = {Findings of the Association for Computational Linguistics: ACL 2023},
pages = {5474-5490},
year = {2023},
month = {07},
address = {Toronto, Canada},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2023.findings-acl.338},
url = {https://aclanthology.org/2023.findings-acl.338},
}
Machel Reid, Mikel Artetxe
Findings of ACL 2023
While prior work has established that the use of parallel data is conducive for cross-lingual learning, it is unclear if the improvements come from the data itself, or if it is the modeling of parallel interactions that matters. Exploring this, we examine the usage of unsupervised machine translation to generate synthetic parallel data, and compare it to supervised machine translation and gold parallel data. We find that even model generated parallel data can be useful for downstream tasks, in both a general setting (continued pretraining) as well as the task-specific setting (translate-train), although our best results are still obtained using real parallel data. Our findings suggest that existing multilingual models do not exploit the full potential of monolingual data, and prompt the community to reconsider the traditional categorization of cross-lingual learning approaches.
@inproceedings{reid2023role
title = {On the Role of Parallel Data in Cross-lingual Transfer Learning},
author = {Reid, Machel and Artetxe, Mikel},
booktitle = {Findings of the Association for Computational Linguistics: ACL 2023},
pages = {5999-6006},
year = {2023},
month = {07},
address = {Toronto, Canada},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2023.findings-acl.372},
url = {https://aclanthology.org/2023.findings-acl.372},
}
Mikel Artetxe, Shruti Bhosale, Naman Goyal, Todor Mihaylov, Myle Ott, Sam Shleifer, Xi Victoria Lin, Jingfei Du, Srinivasan Iyer, Ramakanth Pasunuru, Giridharan Anantharaman, Xian Li, Shuohui Chen, Halil Akin, Mandeep Baines, Louis Martin, Xing Zhou, Punit Singh Koura, Brian O’Horo, Jeffrey Wang, Luke Zettlemoyer, Mona Diab, Zornitsa Kozareva, Veselin Stoyanov
EMNLP 2022
Mixture of Experts layers (MoEs) enable efficient scaling of language models through conditional computation. This paper presents a detailed empirical study of how autoregressive MoE language models scale in comparison with dense models in a wide range of settings: in- and out-of-domain language modeling, zero- and few-shot priming, and full-shot fine-tuning. With the exception of fine-tuning, we find MoEs to be substantially more compute efficient. At more modest training budgets, MoEs can match the performance of dense models using ~4 times less compute. This gap narrows at scale, but our largest MoE model (1.1T parameters) consistently outperforms a compute-equivalent dense model (6.7B parameters). Overall, this performance gap varies greatly across tasks and domains, suggesting that MoE and dense models generalize differently in ways that are worthy of future study. We make our code and models publicly available for research use.
@inproceedings{artetxe2022efficient
title = {Efficient Large Scale Language Modeling with Mixtures of Experts},
author = {Artetxe, Mikel and Bhosale, Shruti and Goyal, Naman and Mihaylov, Todor and Ott, Myle and Shleifer, Sam and Lin, Xi Victoria and Du, Jingfei and Iyer, Srinivasan and Pasunuru, Ramakanth and Anantharaman, Giridharan and Li, Xian and Chen, Shuohui and Akin, Halil and Baines, Mandeep and Martin, Louis and Zhou, Xing and Koura, Punit Singh and O’Horo, Brian and Wang, Jeffrey and Zettlemoyer, Luke and Diab, Mona and Kozareva, Zornitsa and Stoyanov, Veselin},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {11699-11732},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.804},
url = {https://aclanthology.org/2022.emnlp-main.804},
}
Mikel Artetxe, Itziar Aldabe, Rodrigo Agerri, Olatz Perez-de-Viñaspre, Aitor Soroa
EMNLP 2022
The vast majority of non-English corpora are derived from automatically filtered versions of CommonCrawl. While prior work has identified major issues on the quality of these datasets (Kreutzer et al., 2021), it is not clear how this impacts downstream performance. Taking representation learning in Basque as a case study, we explore tailored crawling (manually identifying and scraping websites with high-quality content) as an alternative to filtering CommonCrawl. Our new corpus, called EusCrawl, is similar in size to the Basque portion of popular multilingual corpora like CC100 and mC4, yet it has a much higher quality according to native annotators. For instance, 66% of documents are rated as high-quality for EusCrawl, in contrast with \textless33% for both mC4 and CC100. Nevertheless, we obtain similar results on downstream NLU tasks regardless of the corpus used for pre-training. Our work suggests that NLU performance in low-resource languages is not primarily constrained by the quality of the data, and other factors like corpus size and domain coverage can play a more important role.
@inproceedings{artetxe2022corpus
title = {Does Corpus Quality Really Matter for Low-Resource Languages?},
author = {Artetxe, Mikel and Aldabe, Itziar and Agerri, Rodrigo and Perez-de-Viñaspre, Olatz and Soroa, Aitor},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {7383-7390},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.499},
url = {https://aclanthology.org/2022.emnlp-main.499},
}
Mozes van de Kar, Mengzhou Xia, Danqi Chen, Mikel Artetxe
EMNLP 2022
Masked language models like BERT can perform text classification in a zero-shot fashion by reformulating downstream tasks as text infilling. However, this approach is highly sensitive to the template used to prompt the model, yet practitioners are blind when designing them in strict zero-shot settings. In this paper, we propose an alternative mining-based approach for zero-shot learning. Instead of prompting language models, we use regular expressions to mine labeled examples from unlabeled corpora, which can optionally be filtered through prompting, and used to finetune a pretrained model. Our method is more flexible and interpretable than prompting, and outperforms it on a wide range of tasks when using comparable templates. Our results suggest that the success of prompting can partly be explained by the model being exposed to similar examples during pretraining, which can be directly retrieved through regular expressions.
@inproceedings{vandekar2022dont
title = {Don’t Prompt, Search! Mining-based Zero-Shot Learning with Language Models},
author = {van de Kar, Mozes and Xia, Mengzhou and Chen, Danqi and Artetxe, Mikel},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {7508-7520},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.509},
url = {https://aclanthology.org/2022.emnlp-main.509},
}
Christos Baziotis, Mikel Artetxe, James Cross, Shruti Bhosale
EMNLP 2022
Multilingual machine translation suffers from negative interference across languages. A common solution is to relax parameter sharing with language-specific modules like adapters. However, adapters of related languages are unable to transfer information, and their total number of parameters becomes prohibitively expensive as the number of languages grows. In this work, we overcome these drawbacks using hyper-adapters – hyper-networks that generate adapters from language and layer embeddings. While past work had poor results when scaling hyper-networks, we propose a rescaling fix that significantly improves convergence and enables training larger hyper-networks. We find that hyper-adapters are more parameter efficient than regular adapters, reaching the same performance with up to 12 times less parameters. When using the same number of parameters and FLOPS, our approach consistently outperforms regular adapters. Also, hyper-adapters converge faster than alternative approaches and scale better than regular dense networks. Our analysis shows that hyper-adapters learn to encode language relatedness, enabling positive transfer across languages.
@inproceedings{baziotis2022multilingual
title = {Multilingual Machine Translation with Hyper-Adapters},
author = {Baziotis, Christos and Artetxe, Mikel and Cross, James and Bhosale, Shruti},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {1170-1185},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.77},
url = {https://aclanthology.org/2022.emnlp-main.77},
}
Mengzhou Xia, Mikel Artetxe, Jingfei Du, Danqi Chen, Veselin Stoyanov
EMNLP 2022
Pre-trained masked language models successfully perform few-shot learning by formulating downstream tasks as text infilling. How- ever, as a strong alternative in full-shot settings, discriminative pre-trained models like ELECTRA do not fit into the paradigm. In this work, we adapt prompt-based few-shot learning to ELECTRA and show that it outperforms masked language models in a wide range of tasks. ELECTRA is pre-trained to distinguish if a token is generated or original. We naturally extend that to prompt-based few-shot learning by training to score the originality of the target options without introducing new parameters. Our method can be easily adapted to tasks involving multi-token predictions without extra computation overhead. Analysis shows that ELECTRA learns distributions that align better with downstream tasks.
@inproceedings{xia2022prompting
title = {Prompting ELECTRA: Few-Shot Learning with Discriminative Pre-Trained Models},
author = {Xia, Mengzhou and Artetxe, Mikel and Du, Jingfei and Chen, Danqi and Stoyanov, Veselin},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {11351-11361},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.780},
url = {https://aclanthology.org/2022.emnlp-main.780},
}
Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, Luke Zettlemoyer
EMNLP 2022
Large language models (LMs) are able to in-context learn—perform a new task via inference alone by conditioning on a few input-label pairs (demonstrations) and making predictions for new inputs. However, there has been little understanding of how the model learns and which aspects of the demonstrations contribute to end task performance. In this paper, we show that ground truth demonstrations are in fact not required—randomly replacing labels in the demonstrations barely hurts performance on a range of classification and multi-choce tasks, consistently over 12 different models including GPT-3. Instead, we find that other aspects of the demonstrations are the key drivers of endtask performance, including the fact that they provide a few examples of (1) the label space, (2) the distribution of the input text, and (3) the overall format of the sequence. Together, our analysis provides a new way of understanding how and why in-context learning works, while opening up new questions about how much can be learned from large language models through inference alone.
@inproceedings{min2022rethinking
title = {Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?},
author = {Min, Sewon and Lyu, Xinxi and Holtzman, Ari and Artetxe, Mikel and Lewis, Mike and Hajishirzi, Hannaneh and Zettlemoyer, Luke},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {11048-11064},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.759},
url = {https://aclanthology.org/2022.emnlp-main.759},
}
Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O’Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li
EMNLP 2022
Large-scale generative language models such as GPT-3 are competitive few-shot learners. While these models are known to be able to jointly represent many different languages, their training data is dominated by English, potentially limiting their cross-lingual generalization. In this work, we train multilingual generative language models on a corpus covering a diverse set of languages, and study their few- and zero-shot learning capabilities in a wide range of tasks. Our largest model with 7.5 billion parameters sets new state of the art in few-shot learning in more than 20 representative languages, outperforming GPT-3 of comparable size in multilingual commonsense reasoning (with +7.4% absolute accuracy improvement in 0-shot settings and +9.4% in 4-shot settings) and natural language inference (+5.4% in each of 0-shot and 4-shot settings). On the FLORES-101 machine translation benchmark, our model outperforms GPT-3 on 171 out of 182 directions with 32 training examples, while surpassing the official supervised baseline in 45 directions. We conduct an in-depth analysis of different multilingual prompting approaches, showing in particular that strong few-shot learning performance across languages can be achieved via cross-lingual transfer through both templates and demonstration examples.
@inproceedings{lin2022fewshot
title = {Few-shot Learning with Multilingual Generative Language Models},
author = {Lin, Xi Victoria and Mihaylov, Todor and Artetxe, Mikel and Wang, Tianlu and Chen, Shuohui and Simig, Daniel and Ott, Myle and Goyal, Naman and Bhosale, Shruti and Du, Jingfei and Pasunuru, Ramakanth and Shleifer, Sam and Koura, Punit Singh and Chaudhary, Vishrav and O’Horo, Brian and Wang, Jeff and Zettlemoyer, Luke and Kozareva, Zornitsa and Diab, Mona and Stoyanov, Veselin and Li, Xian},
booktitle = {Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
pages = {9019-9052},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.emnlp-main.616},
url = {https://aclanthology.org/2022.emnlp-main.616},
}
Mikel Artetxe, Jingfei Du, Naman Goyal, Luke Zettlemoyer, Veselin Stoyanov
Findings of EMNLP 2022
Prior work on language model pre-training has explored different architectures and learning objectives, but differences in data, hyperparameters and evaluation make a principled comparison difficult. In this work, we focus on bidirectionality as a key factor that differentiates existing approaches, and present a comprehensive study of its role in next token prediction, text infilling, zero-shot priming and fine-tuning. We propose a new framework that generalizes prior approaches, including fully unidirectional models like GPT, fully bidirectional models like BERT, and hybrid models like CM3 and prefix LM. Our framework distinguishes between two notions of bidirectionality (bidirectional context and bidirectional attention) and allows us to control each of them separately. We find that the optimal configuration is largely application-dependent (e.g., bidirectional attention is beneficial for fine-tuning and infilling, but harmful for next token prediction and zero-shot priming). We train models with up to 6.7B parameters, and find differences to remain consistent at scale. While prior work on scaling has focused on left-to-right autoregressive models, our results suggest that this approach comes with some trade-offs, and it might be worthwhile to develop very large bidirectional models.
@inproceedings{artetxe2022role
title = {On the Role of Bidirectionality in Language Model Pre-Training},
author = {Artetxe, Mikel and Du, Jingfei and Goyal, Naman and Zettlemoyer, Luke and Stoyanov, Veselin},
booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2022},
pages = {3973-3985},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.findings-emnlp.293},
url = {https://aclanthology.org/2022.findings-emnlp.293},
}
Aitor Ormazabal, Mikel Artetxe, Manex Agirrezabal, Aitor Soroa, Eneko Agirre
Findings of EMNLP 2022
Formal verse poetry imposes strict constraints on the meter and rhyme scheme of poems. Most prior work on generating this type of poetry uses existing poems for supervision, which are difficult to obtain for most languages and poetic forms. In this work, we propose an unsupervised approach to generate poems that follow any given meter and rhyme scheme, without requiring any poetic text for training. Our method works by splitting a regular, non-poetic corpus into phrases, prepending control codes that describe the length and end rhyme of each phrase, and training a transformer language model in the augmented corpus. The transformer learns to link the structure descriptor with the control codes to the number of lines, their length and their end rhyme. During inference, we build control codes for the desired meter and rhyme scheme, and condition our language model on them to generate formal verse poetry. Experiments in Spanish and Basque show that our approach is able to generate valid poems, which are often comparable in quality to those written by humans.
@inproceedings{ormazabal2022poelm
title = {PoeLM: A Meter- and Rhyme-Controllable Language Model for Unsupervised Poetry Generation},
author = {Ormazabal, Aitor and Artetxe, Mikel and Agirrezabal, Manex and Soroa, Aitor and Agirre, Eneko},
booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2022},
pages = {3655-3670},
year = {2022},
month = {12},
address = {Abu Dhabi, United Arab Emirates},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.findings-emnlp.268},
url = {https://aclanthology.org/2022.findings-emnlp.268},
}
Jonas Pfeiffer, Naman Goyal, Xi Victoria Lin, Xian Li, James Cross, Sebastian Riedel, Mikel Artetxe
NAACL 2022
Multilingual pre-trained models are known to suffer from the curse of multilinguality, which causes per-language performance to drop as they cover more languages. We address this issue by introducing language-specific modules, which allows us to grow the total capacity of the model, while keeping the total number of trainable parameters per language constant. In contrast with prior work that learns language-specific components post-hoc, we pre-train the modules of our Cross-lingual Modular (X-Mod) models from the start. Our experiments on natural language inference, named entity recognition and question answering show that our approach not only mitigates the negative interference between languages, but also enables positive transfer, resulting in improved monolingual and cross-lingual performance. Furthermore, our approach enables adding languages post-hoc with no measurable drop in performance, no longer limiting the model usage to the set of pre-trained languages.
@inproceedings{pfeiffer2022lifting
title = {Lifting the Curse of Multilinguality by Pre-training Modular Transformers},
author = {Pfeiffer, Jonas and Goyal, Naman and Lin, Xi Victoria and Li, Xian and Cross, James and Riedel, Sebastian and Artetxe, Mikel},
booktitle = {Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
pages = {3479-3495},
year = {2022},
month = {07},
address = {Seattle, United States},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.naacl-main.255},
url = {https://aclanthology.org/2022.naacl-main.255},
}
Machel Reid, Mikel Artetxe
NAACL 2022
Despite the success of multilingual sequence-to-sequence pretraining, most existing approaches rely on monolingual corpora and do not make use of the strong cross-lingual signal contained in parallel data. In this paper, we present PARADISE (PARAllel &Denoising Integration in SEquence-to-sequence models), which extends the conventional denoising objective used to train these models by (i) replacing words in the noised sequence according to a multilingual dictionary, and (ii) predicting the reference translation according to a parallel corpus instead of recovering the original sequence. Our experiments on machine translation and cross-lingual natural language inference show an average improvement of 2.0 BLEU points and 6.7 accuracy points from integrating parallel data into pretraining, respectively, obtaining results that are competitive with several popular models at a fraction of their computational cost.
@inproceedings{reid2022paradise
title = {PARADISE: Exploiting Parallel Data for Multilingual Sequence-to-Sequence Pretraining},
author = {Reid, Machel and Artetxe, Mikel},
booktitle = {Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
pages = {800-810},
year = {2022},
month = {07},
address = {Seattle, United States},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.naacl-main.58},
url = {https://aclanthology.org/2022.naacl-main.58},
}
Aitor Ormazabal, Mikel Artetxe, Aitor Soroa, Gorka Labaka, Eneko Agirre
ACL 2022
Round-trip Machine Translation (MT) is a popular choice for paraphrase generation, which leverages readily available parallel corpora for supervision. In this paper, we formalize the implicit similarity function induced by this approach, and show that it is susceptible to non-paraphrase pairs sharing a single ambiguous translation. Based on these insights, we design an alternative similarity metric that mitigates this issue by requiring the entire translation distribution to match, and implement a relaxation of it through the Information Bottleneck method. Our approach incorporates an adversarial term into MT training in order to learn representations that encode as much information about the reference translation as possible, while keeping as little information about the input as possible. Paraphrases can be generated by decoding back to the source from this representation, without having to generate pivot translations. In addition to being more principled and efficient than round-trip MT, our approach offers an adjustable parameter to control the fidelity-diversity trade-off, and obtains better results in our experiments.
@inproceedings{ormazabal2022principled
title = {Principled Paraphrase Generation with Parallel Corpora},
author = {Ormazabal, Aitor and Artetxe, Mikel and Soroa, Aitor and Labaka, Gorka and Agirre, Eneko},
booktitle = {Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
pages = {1621-1638},
year = {2022},
month = {05},
address = {Dublin, Ireland},
publisher = {Association for Computational Linguistics},
doi = {10.18653/v1/2022.acl-long.114},
url = {https://aclanthology.org/2022.acl-long.114},
}
Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, Todor Mihaylov, Myle Ott, Sam Shleifer, Kurt Shuster, Daniel Simig, Punit Singh Koura, Anjali Sridhar, Tianlu Wang, Luke Zettlemoyer
arXiv preprint
Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models.
@misc{zhang2022opt
title = {OPT: Open Pre-trained Transformer Language Models},
author = {Zhang, Susan and Roller, Stephen and Goyal, Naman and Artetxe, Mikel and Chen, Moya and Chen, Shuohui and Dewan, Christopher and Diab, Mona and Li, Xian and Lin, Xi Victoria and Mihaylov, Todor and Ott, Myle and Shleifer, Sam and Shuster, Kurt and Simig, Daniel and Koura, Punit Singh and Sridhar, Anjali and Wang, Tianlu and Zettlemoyer, Luke},
year = {2022},
month = {05},
publisher = {arXiv},
doi = {10.48550/arXiv.2205.01068},
url = {https://arxiv.org/abs/2205.01068},
}
Nicola De Cao, Ledell Wu, Kashyap Popat, Mikel Artetxe, Naman Goyal, Mikhail Plekhanov, Luke Zettlemoyer, Nicola Cancedda, Sebastian Riedel, Fabio Petroni
TACL
We present mGENRE, a sequence-to-sequence system for the Multilingual Entity Linking (MEL) problem -- the task of resolving language-specific mentions to a multilingual Knowledge Base (KB). For a mention in a given language, mGENRE predicts the name of the target entity left-to-right, token-by-token in an autoregressive fashion. The autoregressive formulation allows us to effectively cross-encode mention string and entity names to capture more interactions than the standard dot product between mention and entity vectors. It also enables fast search within a large KB even for mentions that do not appear in mention tables and with no need for large-scale vector indices. While prior MEL works use a single representation for each entity, we match against entity names of as many languages as possible, which allows exploiting language connections between source input and target name. Moreover, in a zero-shot setting on languages with no training data at all, mGENRE treats the target language as a latent variable that is marginalized at prediction time. This leads to over 50% improvements in average accuracy. We show the efficacy of our approach through extensive evaluation including experiments on three popular MEL benchmarks where mGENRE establishes new state-of-the-art results. Code and pre-trained models at https://github.com/facebookresearch/GENRE.
@article{decao2022multilingual
title = {Multilingual Autoregressive Entity Linking},
author = {De Cao, Nicola and Wu, Ledell and Popat, Kashyap and Artetxe, Mikel and Goyal, Naman and Plekhanov, Mikhail and Zettlemoyer, Luke and Cancedda, Nicola and Riedel, Sebastian and Petroni, Fabio},
journal = {Transactions of the Association for Computational Linguistics},
volume = {10},
pages = {274-290},
year = {2022},
month = {03},
publisher = {MIT Press},
issn = {2307-387X},
doi = {10.1162/tacl_a_00460},
url = {https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00460/110051/Multilingual-Autoregressive-Entity-Linking},
}
Ping Yu, Mikel Artetxe, Myle Ott, Sam Shleifer, Hongyu Gong, Ves Stoyanov, Xian Li
arXiv preprint
All-MLP architectures have attracted increasing interest as an alternative to attention-based models. In NLP, recent work like gMLP shows that all-MLPs can match Transformers in language modeling, but still lag behind in downstream tasks. In this work, we analyze the limitations of MLPs in expressiveness, and propose sparsely activated MLPs with mixture-of-experts (MoEs) in both feature and input (token) dimensions. Such sparse all-MLPs significantly increase model capacity and expressiveness while keeping the compute constant. We address critical challenges in incorporating conditional computation with two routing strategies. The proposed sparse all-MLP improves language modeling perplexity and obtains up to 2$\times$ improvement in training efficiency compared to both Transformer-based MoEs (GShard, Switch Transformer, Base Layers and HASH Layers) as well as dense Transformers and all-MLPs. Finally, we evaluate its zero-shot in-context learning performance on six downstream tasks, and find that it surpasses Transformer-based MoEs and dense Transformers.
@misc{yu2022efficient
title = {Efficient Language Modeling with Sparse all-MLP},
author = {Yu, Ping and Artetxe, Mikel and Ott, Myle and Shleifer, Sam and Gong, Hongyu and Stoyanov, Ves and Li, Xian},
year = {2022},
month = {03},
publisher = {arXiv},
doi = {10.48550/arXiv.2203.06850},
url = {https://arxiv.org/abs/2203.06850},
}