Devamanyu Hazarika

Research Scientist
Meta Superintelligence Labs,
Llama Research

devamanyu@u.nus.edu

I am a Research Scientist at Meta Superintelligence Labs. I spend my time post-training Llama models and teaching them how to code.

Previously I was a Senior Applied Scientist at Amazon and a founding member of Amazon AGI. I was a core developer of Amazon Nova and co-led the model alignment team. Prior to Amazon, I completed my PhD in Computer Science from National University of Singapore.

My research interests include post-training of LLMs with a focus on alignment. I have also extensively worked in the field of conversational AI -- understanding affect, commonsense and multimodal information in dialog contexts and generating controllable, effective, and empathetic responses. I have published over 40 papers in top-tier AI conferences that include ACL, EMNLP, NAACL, AAAI, etc. and regularly serve in the program committee of these conferences, e.g. Senior AC in ACL '25, AC in ACL, EMNLP '23, '24. I have received multiple awards, such as, Best Reviewer Award (ACL 2021), Research Achievement Award (NUS 2019) and the IEEE CIM Outstanding Paper Award (2021).

News

Moved to Meta GenAI -- Llama Research Post-training.
Senior Area Chair at ACL 2025.
Area Chair at ACL 2024.
2 Papers accepted at EMNLP 2023 -- Instruction Induction (main) and ICL for Dialogs (findings)
Papers at EACL 2023, Sigdial 2023 -- Dialog Augmentation, Subjective Dialogs
Area Chair at EMNLP 2023, ACL 2023, AAAI 2024, SICON 2023, AAAI 2023
Publicity Chair at ACL 2023
Papers at EMNLP 2022, ACL 2022, NAACL 2022 -- Inducer Tuning, Adapters, Style Transfer
Defended PhD Thesis and joined Amazon Alexa AI as Applied Scientist

Affiliations

Selected Publications
Complete List: Google Scholar

CESAR: Automatic Induction of Compositional Instructions for Multi-turn Dialogs

Taha Aksu, Devamanyu Hazarika, Shikib Mehri, Seokhwan Kim, Dilek Hakkani-Tür, Yang Liu, Mahdi Namazifar, EMNLP 2023 (Main)

[Paper]

Using In-Context Learning to Improve Dialogue Safety

Nicholas Meade, Spandana Gella, Devamanyu Hazarika, Prakhar Gupta, Di Jin, Siva Reddy, Yang Liu, Dilek Hakkani-Tür, EMNLP 2023 (Findings)

[Paper]

"What do others think?": Task-Oriented Conversational Modeling with Subjective Knowledge

Chao Zhao, Spandana Gella, Seokhwan Kim, Di Jin, Devamanyu Hazarika, Alexandros Papangelis, Behnam Hedayatnia, Mahdi Namazifar, Yang Liu, Dilek Hakkani-Tur, Sigdial 2023

[Paper]

KILM: Knowledge Injection into Encoder-Decoder Language Models (Oral)

Yan Xu, Mahdi Namazifar, Devamanyu Hazarika, Aishwarya Padmakumar, Yang Liu, Dilek Hakkani-Tür, ACL 2023 (Main)

[Paper]

Selective in-context data augmentation for intent detection using pointwise V-information

Yen-Ting Lin, Alexandros Papangelis, Seokhwan Kim, Sungjin Lee, Devamanyu Hazarika, Mahdi Namazifar, Di Jin, Yang Liu, Dilek Hakkani-Tür, EACL 2023

[Paper]

Inducer-tuning: Connecting Prefix-tuning and Adapter-tuning (Oral)

Yifan Chen*, Devamanyu Hazarika*, Mahdi Namazifar, Yang Liu, Di Jin, Dilek Hakkani-Tür , EMNLP 2022 (Main) *Equal Contribution

[Paper] [Slides]

Empowering Parameter-Efficient Transfer Learning by Recognizing the Kernel

Structure in Attention

Yifan Chen*, Devamanyu Hazarika*, Mahdi Namazifar, Yang Liu, Di Jin, Dilek Hakkani-Tür, NAACL 2022 *Equal Contribution

[Paper]

Analyzing Modality Robustness in Multimodal Sentiment Analysis

Devamanyu Hazarika*, Y Li*, B Cheng, S Zhao, Roger Zimmermann, Soujanya Poria, NAACL 2022 *Equal Contribution

[Paper] [Code] [Slides]

So Different Yet So Alike! Constrained Unsupervised Text Style Transfer (Oral)

Aabhinav Kashyap*, Devamanyu Hazarika*, Min-Yen Kan, Roger Zimmermann, Soujanya Poria, ACL 2022 *Equal Contribution

[Paper] [Code]

Attention Biasing and Context Augmentation for Zero-Shot Control of Encoder-Decoder Transformers for Natural Language Generation

Devamanyu Hazarika, Mahdi Namazifar, and Dilek Hakkani-Tur, AAAI 2022

[Paper] [Arxiv]

Domain Divergences: a Survey and Empirical Analysis

Abhinav Kashyap, Devamanyu Hazarika, Min-Yen Kan, and Roger Zimmermann, NAACL 2021

[Paper]

Methods for Numeracy-Preserving Word Embeddings

D Sundararaman, S Si, V Subramanian, G Wang, D Hazarika, L Carin, EMNLP 2020

[Paper]

Misa: Modality-Invariant and -Specific Representations for Multimodal Sentiment Analysis

D Hazarika, R Zimmermann, S Poria, ACM MM 2020

[Paper] [Code]

Kingdom: Knowledge-Guided Domain Adaptation for Sentiment Analysis

D Ghosal, D Hazarika, A Roy, N Majumder, R Mihalcea, S Poria, ACL 2020

[Paper] [Code]

Towards Multimodal Sarcasm Detection (An Obviously Perfect Paper)

S Castro, D Hazarika, V Pérez-Rosas, R Zimmermann, R Mihalcea, S Poria, ACL 2020

[Paper] [Code]

Contextually Grounded Affective Analysis of Media

D Hazarika, Thesis '21

Recognizing Emotion Cause in Conversations

S Poria, N Majumder, D Hazarika, D Ghosal, R Bhardwaj, S Y B Jian, P Hong, R Ghosh, A Roy, N Chhaya, A Gelbukh, R Mihalcea, Cognitive Computation '21

[Paper] [Code]

Beneath the Tip of the Iceberg: Current Challenges and New Directions in Sentiment Analysis Research

S Poria, D Hazarika, N Majumder, R Mihalcea, IEEE Transactions on Affective Computing '20

[Paper]

Multimodal Research in Vision and Language: A Review of Current and Emerging Trends

S Uppal, S Bhagat, D Hazarika, N Majumder, S Poria, R Zimmermann, A Zadeh, Information Fusion '22

[Paper]

Conversational Transfer Learning for Emotion Recognition

D Hazarika, S Poria, R Zimmermann, R Mihalcea, Information Fusion '21

[Paper] [Code]

Analyzing the Domain Robustness of Pretrained Language Models, Layer by Layer

AR Kashyap, L Mehnaz, B Malik, A Waheed, D Hazarika, M-Y Kan, and R Shah, Adapt-NLP, EACL 2021

[Paper]

Causal Augmentation for Causal Sentence Classification

FA Tan, D Hazarika, S K Ng, S Poria, R Zimmermann, CL+NLP 2021

[Paper] [Code]

Page updated

Google Sites

Report abuse