Ankit Arun

I'm a Machine Learning Engineer in Meta Reality Labs at Meta currently working on Multimodal AI. I previously worked at PatternEx, a cybersecurity startup working with Prof. Kalyan Veeramachaneni from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and Dr. Ignacio Arnaldo on cybersecurity applications of machine learning.

My current work focuses on making natural language generation systems more reliable and data-efficient. I'm particularly interested in developing techniques to improve the factual consistency and faithfulness of abstractive summarization models. I've worked on controlling confounding effects from irrelevant input text and building adaptive acceptability classifiers for NLG quality assessment. I also explore methods like few-shot learning and self-training to reduce the data requirements for deploying production NLG systems.

Prior to my current role, I worked on applying machine learning to cybersecurity problems like malicious domain detection and network log analysis. I developed systems for continuous model updating to handle evolving threats and built platforms to automate the deployment of ML models in production security environments. My research aims to make AI systems more robust, reliable and practical to deploy in real-world applications.

Publications

Improving Faithfulness of Abstractive Summarization by Controlling Confounding Effect of Irrelevant Sentences

Improving Faithfulness of Abstractive Summarization by Controlling Confounding Effect of Irrelevant Sentences

Asish Ghoshal, Arash Einolghozati, A. Arun, Haoran Li, L. Yu, Yashar Mehdad, S. Yih, Asli Celikyilmaz

arXiv.org 2022

Best Practices for Data-Efficient Modeling in NLG:How to Train Production-Ready Neural Models with Less Data

A. Arun, Soumya Batra, Vikas Bhardwaj, Ashwini Challa, Pinar E. Donmez, P. Heidari, Hakan Inan, Shashank Jain, Anuj Kumar, Shawn Mei, Karthika Mohan, Michael White

International Conference on Computational Linguistics 2020

Acquire, adapt, and anticipate: continuous learning to block malicious domains

Acquire, adapt, and anticipate: continuous learning to block malicious domains

Ignacio Arnaldo, A. Arun, Sumeeth Kyathanahalli, K. Veeramachaneni

2018 IEEE International Conference on Big Data (Big Data) 2018

Learning Representations for Log Data in Cybersecurity

Ignacio Arnaldo, Alfredo Cuesta-Infante, A. Arun, Mei Lam, Costas Bassias, K. Veeramachaneni

International Conference on Cyber Security Cryptography and Machine Learning 2017

Building Adaptive Acceptability Classifiers for Neural NLG

Building Adaptive Acceptability Classifiers for Neural NLG

Soumya Batra, Shashank Jain, P. Heidari, A. Arun, Catharine Youngs, Xintong Li, Pinar E. Donmez, Shawn Mei, Shiunzu Kuo, Vikas Bhardwaj, Anuj Kumar, Michael White

Conference on Empirical Methods in Natural Language Processing 2021

Getting to Production with Few-shot Natural Language Generation Models

Getting to Production with Few-shot Natural Language Generation Models

P. Heidari, Arash Einolghozati, Shashank Jain, Soumya Batra, Lee F. Callender, A. Arun, Shawn Mei, Sonal Gupta, Pinar E. Donmez, Vikas Bhardwaj, Anuj Kumar, Michael White

SIGDIAL Conferences 2021

Structure-to-Text Generation with Self-Training, Acceptability Classifiers and Context-Conditioning for the GEM Shared Task

Structure-to-Text Generation with Self-Training, Acceptability Classifiers and Context-Conditioning for the GEM Shared Task

Shreyan Bakshi, Soumya Batra, P. Heidari, A. Arun, Shashank Jain, Michael White

IEEE Games Entertainment Media Conference 2021

Semantic Role Labeling for Process Recognition Questions

Semantic Role Labeling for Process Recognition Questions

Samuel Louvan, Chetan Naik, Veronica E. Lynn, A. Arun, Niranjan Balasubramanian, Peter Clark

Shooting the moving target: machine learning in cybersecurity

Shooting the moving target: machine learning in cybersecurity

A. Arun, Ignacio Arnaldo

USENIX Conference on Operational Machine Learning 2019