MBZUAI at ACL2023

Monday, July 10, 2023

At last count, MBZUAI scientists have 26 papers accepted at the 61st Annual Meeting of the Association for Computational Linguistics (ACL 2023). In total, 11 MBZUAI faculty, researchers, postdocs, and students have academic research accepted at ACL 2023 alongside their national and international co-authors. This year’s ACL meeting is slated to take place in Toronto, Canada from July 9 to July 14, 2023.

As a conference, ACL is one of the three top-ranked conferences in the Natural Language Processing (NLP) field alongside EMNLP and NAACL. MBZUAI ranks 19 in the world, according to CSRankings.org, for publishing in top AI conferences in NLP, AI, machine learning (ML), and computer vision (CV).

Helping LLMs tackle novel tasks

Topping the list of MBZUAI publishers at ACL 2023 is Assistant Professor of Natural Language Processing Alham Fikri Aji, who is co-author of eight papers with more than 80 collaborating scientists from institutions across the globe. One of the papers titled: “Crosslingual Generalization through Multitask Finetuning”, explores how large language models can be improved in their ability to encounter new tasks in languages other than English, using fine-tuning methods. Labeled multitask prompted finetuning (MTF), the team improve performance on multilingual tasks leading to various state-of-the-art zero-shot results.

In the paper, they ultimately find that models are capable of dealing with novel tasks in languages that they have never intentionally seen, and they conjecture that the models are learning higher-level capabilities that are both task and language agnostic. Their code, datasets and models are freely available at https://github.com/ bigscience-workshop/xmtf

Four of Aji’s papers have been featured by Bloomberg in its Tech at Bloomberg Select Research Papers 2023 list. The list also includes Aji and Acting Provost, Department Chair and Professor of Natural Language Processing Timothy Baldwin’s award-winning EACL paperNusaX: Multilingual Parallel Sentiment Dataset for 10 Indonesian Local Languages”.

Recognizing a colleague

Deputy Department Chair and Professor of Natural Language Processing Preslav Nakov published five papers with his co-authors. In particular, Nakov co-authored a research paper titled: “bgGLUE: A Bulgarian General Language Understanding Evaluation Benchmark”, which was published in honor of, and with, Yale Computer Scientist Dragomir R. Radev, who passed away in late March, 2023.

“We dedicate this work to the memory of Dragomir Radev, who is a co-author of this paper,” the authors wrote. “He was not only an exceptional computer scientist, but one of the kindest and most humble people many of us have ever known. He cared deeply about Bulgarian NLP and Bulgarian NLP researchers. He was also the one who gave the idea for, and remained the main driving force behind, the Bulgarian GLUE project – he will be greatly missed.”

MBZUAI Faculty featured at ACL 2023

  • Acting Provost, Associate Provost for Academic Affairs, Department Chair of Natural Language Processing, and Professor of Natural Language Processing Timothy Baldwin
  • Deputy Department Chair of Natural Language Processing, and Professor of Natural Language Processing Preslav Nakov
  • Assistant Professor of Natural Language Processing Alham Fikri Aji
  • Assistant Professor of Natural Language Processing Ekaterina Kochmar
  • Visiting Associate Professor of Natural Language Processing and Machine Learning Muhammad Abdul-Mageed
  • Visiting Associate Professor of Computer Vision Xiaodan Liang
  • Affiliated Professor of Machine Learning Chih-Jen Lin.

Related

thumbnail
Thursday, January 09, 2025

LLMs 101: Large language models explained

LLMs are a staple of AI, but what exactly are they? Our 101 guide breaks down the.....

  1. tokens ,
  2. LLM360 ,
  3. llms ,
  4. llm ,
  5. large language models ,
  6. nlp ,
  7. open source ,
  8. natural language processing ,
Read More
thumbnail
Monday, January 06, 2025

Accelerating neural network optimization: The power of second-order methods

A team from MBZUAI presented a new approach for optimizing neural networks at the recent NeurIPS conference.

  1. second-order ,
  2. optimization ,
  3. neural networks ,
  4. neurips ,
  5. research ,
  6. students ,
Read More
thumbnail
Wednesday, December 25, 2024

Machine learning 101

From optimal decision making to neural networks, we look at the basics of machine learning and how.....

  1. prediction ,
  2. machine learning ,
  3. research ,
  4. deep learning ,
  5. ML ,
  6. algorithms ,
Read More