I am M Saiful Bari
. You may also call me Maruf
. I am a doctoral candidate at NTU, Singapore. My supervisor is Prof. Dr. Shafiq Joty. My research objective is to develop deep models that have the notion of humanity (brain motivated). I spend a huge amount of time and effort exploring unsupervised training and their potential contribution to the generalizability and distributional shift of the language model (LM). In my early work, I investigated adversarial training and semi-supervised learning to transfer knowledge from a high resource to a low resource language. My recent work largely involves efficient transductive few-shot inference and parameter efficient multitask inference via prompt tuning. At the core of my work, I investigate distribution shifts between training vs inference data and how we can solve this distributional shift using various methods (i.e., semi-supervised learning, multitask discrete prompting, data distillation, model distillation). I love tooling, debugging and training large language models.
B.Sc in Computer Science and Engineering, 2016
Islamic University of Technology
Ph.D. Candidate, 2020
Nanyang Technological University
7 May | Three paper accepted at ACL'23. See you at Toronto. More on this coming soon. |
---|---|
15 March | Check out our recent pre-print xCodeEval: A Large Scale Multilingual Multitask Benchmark for Code Understanding, Generation, Translation and Retrieval paper. |
20 Dec | Check out our recent pre-print SPT: Semi-Parametric Prompt Tuning for Multitask Prompted Learning paper. |
---|---|
18 Dec | Check out our recent pre-print BLOOM+1: Adding Language Support to BLOOM for Zero-Shot Prompting paper. |
26 Nov | Returned from Amazon d2l summer internship. |
3 Nov | Check out our recent pre-print Crosslingual Generalization through Multitask Finetuning paper. |
1-Oct | Our paper (What Language Model to Train if You Have One Million GPU Hours?) got accepted at EMNLP'22 findings. |
05-Jul | Joined Amazon d2l team as a summer intern. |
11-Apr | Check out our new ACL'22 workshop paper What Language Model to Train if You Have One Million GPU Hours?. |
7-Feb | Check out our pre-print PromptSource paper. |
2-Feb | Check out my recent talk on T0++ paper. |
19-Nov | Our paper T0++ got accepted as a spotlight paper at ICLR'22 |
---|---|
15-Oct | Our paper Multitask Prompted Training Enables Zero-Shot Task Generalization is online. Model (T0++) , Dataset (P3) |
22-Sept | Check out my Recent talk on Finetuned Language Models Are Zero-Shot Learners at NTU-NLP lab. |
25-Aug | My internship work from Amazon got accepted as a short paper in EMNLP. |
3-May | Two paper UXLA (in main conference) and AugVic (in findings) got accepted in ACL-IJCNLP-2021 . |
6-Nov | Get back to the PhD study after completing my internship with Amazon Lex Team. |
---|---|
15-Oct | Our paper LNMap is accepted in EMNLP-2020. |
03-Aug | I will be starting my internship at Amazon, with Lex Team. |
28-Aug | Passed my PhD Qualifying Examination (QE). |
28-Apr | Preprint of our new paper LNMap released in Arxiv. |
24-Apr | Preprint of our new paper MultiMix released in Arxiv. |
25-Mar | Gave a talk on mBART (paper ) in NTU-NLP. |
13-Feb | Presenting our Cross-lingual-NER paper on AAAI-2020. |
4-Jan | Got AAAI-2020 travel scholarship. See you in New York. |
9-Nov | I will be a Teaching Assistant, for Graduate Deep Learning Course of NLP , NTU |
---|---|
11-Nov | A paper accepted on AAAI 2020. |
15-Jun | A paper accepted on ACL 2019. |
25-Jan | Shows MT system to industry and government stakeholders. |
15-Jan | Joins NTU-NLP lab as a PhD student. |
Language
Language
C++ implementation of various algorithms.
This is a tool to translate an English sentence into Malay and vice versa.
Finetuned Language Models Are Zero-Shot Learners
.