Arnav Goel

Hi! I am Arnav, a final year undergraduate at IIIT Delhi, where I am pursuing a B.Tech in Computer Science and Artificial Intelligence. Currently, I am working as a Research Intern at the INK-Lab, under the guidance of Dr. Xiang Ren. In parallel, I serve as a Research Assistant at McGill University and Mila-Quebec AI Institute, supervised by Dr. Jackie Chi Kit Cheung. My work here includes a collaboration with the FATE Team at Microsoft Research Montreal. In the past, I have been fortunate to work under Dr. Sameep Mehta and Dr. Nishtha Madaan from IBM Research India and Dr. Anubha Gupta and Dr. Rajiv Ratn Shah at IIIT Delhi.

Research

My research interests span large language models (LLMs), machine learning, and speech processing. I am currently working on enhancing the transparency, trustworthiness and fairness of large-scale foundation models, with a particular focus on understanding various influences driving their memorization and generalization capabilities. Additionally, I have previously worked on: (1) post-processing solutions to adapt LLMs to enterprise settings, ensuring their safety and robustness; (2) novel architectures and improved representation learning to adapt prosody generation and speech recognition systems to long-tail speaker distributions; (3) measuring representation harms like erasure in NLP systems through a novel conceptualization; and (4) improving LLMs and vision-language models (VLMs) on complex question-answering and reasoning tasks. A full list of my publications is available at this link.

Outside of research, I am a big-time history buff and love spending time watching documentaries. I have recently developed a strong penchant for cooking and enjoy experimenting with new recipes in my free time. Swimming has always been my go-to activity for relaxation, and I am always up for an intense game of squash :)

Feel free to reach out if you would like to chat about large language models (LLMs), explainability, data or machine learning in general. I am also always happy to offer advice to high-schoolers or undergrads exploring their way into AI research. Don’t hesitate to drop me an email if you think I can help!

News

January 2025: Selected for the Google Deepmind Research Symposium 2025. Looking forward to connect with people in Bangalore!
December 2024: Paper titled “Optimizing Multimodal Large Language Models for Scientific VQA through Caption-Aware Supervised Training” accepted at the AI4EDU-Workshop @ AAAI-2025! This work was done as part of my research at the MIDAS-Lab @ IIIT-Delhi.
October 2024: My team DBkaScam was placed 6th amongst 20000+ teams in the prestigious Amazon ML Challenge 2024! Check out our winning solution at this link
June 2024: Full paper titled “Exploring Multilingual Unseen Speaker Emotion Recognition: Leveraging Co-Attention Cues in Multitask Learning” accepted at INTERSPEECH 2024. This work was co-led by me as part of my undergraduate thesis at IIIT-Delhi.
May 2024: Thrilled to share that I have started my internship at the INK-Lab in the USC NLP Group as one of the 15 IUSSTF-Viterbi Research Scholars selected from all over India. Looking forward to an incredible summer in Los Angeles :)
February 2024: 2 papers on speech processing and multilingual prosody transfer accepted at the Tiny Track at ICLR 2024!
November 2023: Paper titled “LLMGuard: Guarding Against Unsafe LLM Behavior” accepted at the Demonstrations Track at AAAI 2024. This work was done as part of my GRM internship at IBM Research India.
August 2023: 2 papers on improving LLMs in scientific-question answering through RAG and alignment were accepted at BDA-2023. This work was done as part of my research for the Summer Undergraduate Research Fellowship (SURF) at IIIT-Delhi.