I am a final year Computer Science PhD student in the MURGe Lab (part of the bigger UNC NLP lab) at the University of North Carolina at Chapel Hill, advised by Prof. Mohit Bansal.

My research goal is to understand the role of data in the training of deep learning models. How to select (data pruning), order (curriculum learning) and modify (data augmentation) training data to achieve better performance in deep learning models - are some of the questions I have explored in my work. Previously, I have worked on a diverse set of topics – (1) story visualization, (2) causal debiasing, (3) ML for public health, and (4) biomedical NLP (information extraction).

During my PhD, I have spent three wonderful summers, interning at Snap Research, AI2 Prior and Adobe Research. Before starting my PhD, I was a Data Scientist at Sciome building information extraction pipelines for biomedical research articles. I did my M.S. in Biomedical & Health Information from University of Washington, Seattle, where I worked on ML solutions for public health problems at the IHME. I completed my undergrad at IIT Kharagpur, India.

Recent News

• February 2024 : Work on evaluation of conversational memory in LLM agents is out on arxiv. See a summary here.
• February 2024 : Work on cross-task inconsistency is accepted to TMLR! Explore the project here.
• October 2023 : Work on data pruning is available on arxiv; now accepted to ICLR 2024!
• October 2023 : Paper on causal debiasing using information minimization accepted to Findings of EMNLP 2023. Also to be presented at NeurIPS 2023 Causal Representation Learning Workshop.
• April 2023 : Our work done in collaboration with AI2 on cross-task inconsistency is out on arxiv. Explore the probing dataset.
• August 2022 : Our paper on StoryDALL-E is published in ECCV 2022 and available for demo!
• April 2022 : Two new papers on curriculum learning (main track) and video-based multimodal intent discovery (findings) accepted to NAACL!
• March 2022 : Our paper on contextual code search has been accepted to DL4C workshop at ICLR 2022! Explore and download openly available tool extension here.
• August 2021 : Our paper on structured inputs for story visualization has been accepted to EMNLP 2021 for an oral presentation!
• July 2021 : Recognized as Outstanding Reviewer for ACL-IJCNLP 2021
• May 2021 : Our paper on code-generation from structured inputs has been accepted to Findings of ACL 2021!
• March 2021 : Gave an invited talk at the ICLR SEDL Workshop on ML solutions for public health!
• February 2021 : Our paper on story visualization has been accepted to NAACL 2021! (paper, code)
• December 2020 : Published a preprint on the use of Technology in COVID-Response in African; now accepted to Scientific African. (paper)
• September 2020 : Our research paper "Adversarial Augmentation Policy Search for Domain and Cross-Lingual Generalization in Reading Comprehension" is accepted to the Findings of EMNLP, 2020. (paper, code)
• July 2019: Paper accepted to Journal of the American Medical Informatics Association (JAMIA) Open.
• August 2018:Paper accepted to Journal of the American Medical Association (JAMA) Open and covered by leading news outlets.

Publications [Google Scholar] [dblp]

Data Curation

• D2 Pruning: Message Passing for Balancing Diversity and Difficulty in Data Pruning
  Adyasha Maharana, Prateek Yadav and Mohit Bansal
  [paper, code]

• Exposing and addressing cross-task inconsistency in unified vision-language models
  Adyasha Maharana, Amita Kamath, Christopher Clark, Mohit Bansal and Aniruddha Kembhavi
  [paper, code, dataset]

• On Curriculum Learning for Commonsense Reasoning
  Adyasha Maharana and Mohit Bansal
  NAACL 2022 [paper, code]

• Adversarial Augmentation Policy Search for Domain and Cross-Lingual Generalization in Reading Comprehension
  Adyasha Maharana and Mohit Bansal
  Findings of EMNLP 2020 [paper, code]

Story Visualization

• StoryDALL-E: Adapting Pretrained Text-to-Image Transformers for Story Continuation
  Adyasha Maharana and Mohit Bansal
  ECCV 2022 [paper, code, demo]

• Integrating Visuospatial, Linguistic and Commonsense Structure into Story Visualization (Oral)
  Adyasha Maharana and Mohit Bansal
  EMNLP 2021 [paper, code]

• Improving Generation and Evaluation of Visual Stories via Semantic Consistency
  Adyasha Maharana, Darryl Hannan and Mohit Bansal
  NAACL 2021 [paper, code]

Miscellaneous Methods

• Debiasing Multimodal Models via Causal Information Minimization
  Vaidehi Patil, Adyasha Maharana and Mohit Bansal
  EMNLP 2023 (Findings) [paper, code]

• Multimodal Intent Discovery from Livestream Videos
  Adyasha Maharana, Quan Hung Tran, Franck Dernoncourt, Seunghyun Yoon, Trung Bui, Walter Chang and Mohit Bansal
  NAACL 2022 (Findings) [paper, code]

• Analysis of Tree-based Input and Architecture for Code Generation
  Samip Dahal, Adyasha Maharana, and Mohit Bansal
  Findings of ACL 2021 [paper, code]

Applications

• Use of Technology and Innovations in the COVID-19 Pandemic Response in Africa
  Adyasha Maharana, Morine Amutorine, Moinina David Sengeh and Elaine O. Nsoesie
  Preprint [paper]

• Detecting reports of unsafe foods in consumer product reviews
  Adyasha Maharana, Kunlin Cai, Joseph Hellerstein, Yulin Hswen, Michael Munsell, Valentina Staneva, Miki Verma, Cynthia Vint, Derry Wijaya and Elaine O Nsoesie
  Journal of the American Medical Informatics Association (JAMIA Open), 2019 [paper]

• Quantifying the Impact of the Built Environment on Neighborhood Crime Rates
  Adyasha Maharana, Quynh C. Nguyen, and Elaine O. Nsoesie
  AI for Social Good Workshop (co-located with ICML), 2019 [paper]

• A Pragmatic Approach to Information Extraction for Systematic Review
  Adyasha Maharana, Arpit Tandon, Eric Wimberley, Mihir Shah, Ruchir Shah and Brian E. Howard
  Text Analysis Conference (TAC), 2018 [paper]

• Use of deep learning to examine the association of the built environment with prevalence of neighborhood adult obesity
  Adyasha Maharana and Elaine Nsoesie
  Journal of the American Medical Association (JAMA Network Open), 2018 [paper, code]

• Clinical Event Detection with Hybrid Neural Architecture
  Adyasha Maharana, Meliha Yetisgen
  BioNLP (co-located with ACL), 2017 [paper]

• Social media as a sentinel for disease surveillance: what does sociodemographic status have to do with it?
  Elaine O. Nsoesie, Luisa Flor, Jared Hawkins, Adyasha Maharana, Tobi Skotnes, Fatima Marinho, and John S. Brownstein
  PLOS Currents Outbreaks 2016 [paper, code]

Teaching

• Digital Signal Processing Laboratory - Spring 2015
  

• Micro-controllers & Microprocessors Laboratory - Fall 2014
  

Professional Service

• Reviewer for JAMA Network Open, Frontiers in Artificial Intelligence, RobustML workshop at ICLR 2021, ACL 2020, EMNLP 2021, ACL 2021, NAACL 2021 and ACL Rolling Review
• Member of Program Committee for Causual Inference + NLP Workshop at EMNLP 2021