Koustuv Sinha
McGill University / Montreal Institute of Learning Algorithms.
Room 108,
McConnell Engineering Bldg
3480 University St.
Montreal, Quebec H3A OE9
My research interests lies in the intersection of Generative Language Modeling (Dialog Systems) and Neural Reasoning in Discrete domain (language and graphs). I am also highly interested in exploring linguistic and logical priors in designing machine learning systems which are robust and systematic.
Currently, I live in the wonderful city of Montreal, Canada, being a PhD candidate at McGill University in the School of Computer Science, supervised by Joelle Pineau and William L. Hamilton. I am also a Research PhD intern at Facebook AI Research (FAIR), Montreal. Previously, I worked at Indian Institute of Technology, Kanpur at the position of Senior Project Associate, as well as NRI Fintech, Kolkata, as Front End Developer. You can find my detailed CV here
news
| Apr 15, 2021 | Announcing the pre-print of our paper “Masked Language Modeling and the Distributional Hypothesis: Order Word Matters Pre-training for Little”. We find RoBERTa trained with sentence word order shuffled data performs remarkably close to natural word order pre-trained models on several downstream and probing tasks! |
| Jan 1, 2021 | Excited to announce the public release of a pre-print of our paper “Unnatural Language Inference”, where we stumble upon the weird language understanding mechanisms employed by NLU models! |
| Oct 2, 2020 | Happy to announce our paper “Measuring Systematic Generalization in Neural Proof Generation with Transformers” is accepted at NeurIPS 2020! |
| Sep 5, 2020 | Excited to announce the 2020 edition of the ML Reproducibility Challenge! We now cover 7 major ML conferences, do check it out! |
| Aug 5, 2020 | We released a new blog post on ML Reproducibility Tools and Best Practices. Check it out! |