I am a PhD student in Computer Science at Mila and McGill where I am supervised by Prof. Dzmitry Bahdanau and Prof. Siva Reddy. Previously, I spent 2.5 amazing years as a Research Fellow at Microsoft Research India, where I worked with Dr. Navin Goyal. I also interned with the AllenNLP team at the Allen Institute for Artificial Intelligence (AI2). At AI2, I worked with Pradeep Dasigi on evaluating code generation in LLMs.

I do research in Natural Language Processing. My work focuses on analyzing neural models to
(1) understand their abilities and limitations towards modeling various aspects of language, and
(2) understand the underlying factors responsible for their behavior.
My hope is that the knowledge derived from such analysis works will eventually help design robust and interpretable systems that exhibit a human-like understanding of language.

Keywords: in-context learning, compositional generalization, analysis and interpretability, evaluation

I graduated with B.E. (Hons.) in Computer Science from BITS Pilani - Goa Campus, India in 2020. For more details about my background, refer to my CV. If you'd like to chat with me about my work or research in general, feel free to reach out!
News
Apr 24, 2024

New paper on AI safety investigating the transferability of adversarial triggers in LLMs.

Mar 15, 2024

Our paper on evaluating code generation in LLMs has been accepted at NAACL 2024!

Jan 16, 2024

Our paper on understanding in-context learning in Transformers and LLMs has been accepted at ICLR 2024 for oral presentation (top 1.2%)!

May 02, 2023
Publications
  Google Scholar|   Semantic Scholar

Universal Adversarial Triggers Are Not Universal
Nicholas Meade, Arkil Patel, Siva Reddy
Preprint
pdf code abstract

Evaluating In-Context Learning of Libraries for Code Generation
Arkil Patel, Siva Reddy, Dzmitry Bahdanau, Pradeep Dasigi
NAACL'24
pdf code abstract

Understanding In-Context Learning in Transformers and LLMs by Learning to Learn Discrete Functions
Satwik Bhattamishra, Arkil Patel, Phil Blunsom, Varun Kanade
ICLR'24 [Oral]
pdf code abstract

Simplicity Bias in Transformers and their Ability to Learn Sparse Boolean Functions
Satwik Bhattamishra, Arkil Patel, Varun Kanade, Phil Blunsom
ACL'23
pdf code abstract

Revisiting the Compositional Generalization Abilities of Neural Sequence Models
Arkil Patel, Satwik Bhattamishra, Phil Blunsom, Navin Goyal
ACL'22
pdf code abstract

Are NLP Models really able to Solve Simple Math Word Problems?
Arkil Patel, Satwik Bhattamishra, Navin Goyal
NAACL'21
pdf code abstract article

VehicleChain: Blockchain-based Vehicular Data Transmission Scheme for Smart City
Arkil Patel, Naigam Shah, Trupil Limbasiya, Debasis Das
IEEE SMC'19
pdf

Service
Teaching
  • Winter 2024: Teaching Assistant for COMP 596: From Natural Language to Data Science - McGill University
  • Winter 2023: Teaching Assistant for COMP 596: From Natural Language to Data Science - McGill University
  • Winter 2020: Teaching Assistant for BITS F312: Neural Networks and Fuzzy Logic - BITS Goa
  • Winter 2019: Teaching Assistant for CS F415: Data Mining - BITS Goa
Reviewer     COLM 2024   ACL Rolling Review   ACL 2023   EMNLP 2021, 2022, 2023   NAACL 2021   AAAI 2022

BITS Pilani
2016 - 2020
Microsoft Research India
2019 - 2022
Allen Institute for AI
Summer 2023
Mila - Quebec AI Institute
2022 - Present
McGill University
2022 - Present
  Template: Sebastin