Dhanajit Brahma

I am a postdoctoral associate at Duke University, where I work under the supervision of Prof. Ricardo Henao. I completed my Ph.D. in Computer Science and Engineering at IIT Kanpur under the guidance of Prof. Piyush Rai. I have also worked as a research consultant at KAUST. My current research interests are developing frameworks for causal inference, test-time adaptation, agentic memory systems, and agentic AI.

Scholar  /  DBLP  /  Github

News

• May 2026 Our paper "SAGE: A Novelty Gate for Efficient Memory Evolution in Agentic LLMs" is now available on arXiv. Code is available on GitHub.
• Apr 2026 Our paper, "Calibrating Model-Based Evaluation Metrics for Summarization" has been accepted by ACL 2026.
• Jan 2026 Our paper, "Neural Doubly Robust Proximal Causal Estimation" has been accepted by AISTATS 2026.
• Dec 2025 I presented our paper "Coupling Generative Modeling and an Autoencoder with the Causal Bridge" at NeurIPS 2025, San Diego, CA, US.
• Sep 2025 Our paper, "Coupling Generative Modeling and an Autoencoder with the Causal Bridge" has been accepted by NeurIPS 2025.
• Oct 2024 Joined Duke University as a Postdoctoral Associate working under Prof. Ricardo Henao.
• Feb 2024 Invited to ACM India ARCS/Annual Event 2024 at NISER, Bhubaneswar, to present our paper titled "A Probabilistic Framework for Lifelong Test-Time Adaptation".
• Jun 2023 I presented our paper "A Probabilistic Framework for Lifelong Test-Time Adaptation" at CVPR 2023, Vancouver, Canada.
• May 2023 I received Google Travel Grant for presenting our paper "A Probabilistic Framework for Lifelong Test-Time Adaptation" at CVPR 2023.
• Feb 2023 Our paper, "A Probabilistic Framework for Lifelong Test-Time Adaptation" has been accepted by CVPR 2023.
• Oct 2021 Our paper titled "Fine-Grained Emotion Prediction by Modeling Emotion Definitions" received the Best Student Paper award in ACII 2021.
• Aug 2021 I presented our work virtually at the CSSL workshop, IJCAI 2021.
• Jul 2021 Our paper, "Hypernetworks for Continual Semi-Supervised Learning" received the Best Student Paper award at the CSSL workshop, IJCAI 2021.
• Jun 2020 Our team (my thesis supervisor and me) is one of the winners of Qualcomm Innovation Fellowship 2020 for the innovation titled "Continual Semi-Supervised Multi-Objective Learning".
• Feb 2020 I presented our work on "Zero-shot learning" in AAAI 2020, New York, USA.
• Nov 2019 Our paper titled "Meta-learning for generalized zero-shot learning" has been accepted by AAAI 2020.
• Nov 2019 Our paper titled "Deep Attentive Ranking Networks for Learning to Order Sentences" has been accepted by AAAI 2020.

Research Interests

• Machine Learning
• Probabilistic Machine Learning
• Causal Inference
• Deep Learning
• Computer Vision
• Natural Language Processing
• Agentic Memory Systems

Publications

Preprints

Conference & Journal Publications

Education

Academic Experience

Teaching Assistant

• CS776: Deep Learning For Computer Vision
• CS771: Introduction to Machine Learning
• CS698X: Topics in Probabilistic Modeling and Inference
• CS647: Advanced Topics in Algorithms and Data Structures
• CS673: Machine Translation
• CS251: Computing Laboratory - I

Reviewer for journals and conferences

• TPAMI — IEEE Transactions on Pattern Analysis and Machine Intelligence — 2021, 2023
• IJCV — International Journal of Computer Vision — 2025
• NeurIPS — Neural Information Processing Systems — 2022, 2023, 2024, 2025
• ICML — International Conference on Machine Learning — 2024, 2025, 2026
• ICLR — International Conference on Learning Representations — 2023, 2024, 2026
• AAAI — Association for the Advancement of Artificial Intelligence — 2021, 2022, 2023, 2024, 2025
• ARR — ACL Rolling Review — Jan 2026
• AISTATS — International Conference on Artificial Intelligence and Statistics — 2026
• CoLLAs — Conference on Lifelong Learning Agents — 2024, 2026
• ACML — Asian Conference on Machine Learning — 2022

Academic Achievements

• ★
  Winner of Qualcomm Innovation Fellowship India 2020 and awarded a research grant for the innovation titled "Continual Semi-Supervised Multi-Objective Learning"
• ★
  Awarded Best Student Paper Award for the paper titled "Hypernetworks for Continual Semi-Supervised Learning" at CSSL@IJCAI 2021
• ★
  One of the top 20 teams in Microsoft AI Challenge India 2018 on Machine Reading Comprehension
• ★
  Awarded Certificate of Academic Distinction for securing maximum CPI among all department students during B.Tech. at IIT Jodhpur for two consecutive academic years, 2014-2015 and 2015-2016

Contact

Email: dhanajit[dot]brahma[at]duke[dot]edu

© 2026 Dhanajit Brahma