[“Aishwarya Maheswaran”]

I am a 5th-year PhD student currently at the Indian Institute of Technology Hyderabad. I work in Natural Language Processing with a focus on Interpretability and methods for controllable text generation.

My research interests lie in utilizing interpretability techniques, such as Representational Engineering (RE) and Mechanistic Interpretability, to understand the inner representations in models and gain insights to control them, thereby making them generate appropriately controlled text. My work includes developing methods to control toxicity in Web search Auto-Completions and understanding how large language models (LLMs) represent emotions and demonstrate empathy. My PhD supervisor at IITH is Dr. Maunendra Sankar Deskar. My Co-guide from Swinburne University is Dr. Caslon Chua.

Updates

2026: Paper Accepted: A Unified View on Emotion Representation in Large Language Models at EACL Main Conference EACL!
2025: Paper Accepted: Probing the Inherent Ability of Large Language Models for Generating Empathetic Responses at IEEE Swiss Conference on Data Science, SDS.
2024: Grateful to receive travel grants from ACM/IARCS and PAKDD conference to attend PAKDD.
2024: Paper Accepted:DAC: Quantized Optimal Transport Reward-based Reinforcement Learning Approach to Detoxify Query Auto-Completion at SIGIR.
2024: Paper Accepted: DQAC: Detoxifying Query Auto-completion with Adapters at PAKDD.
2024: Visited Swinburne University, Melbourne for roughly a year as part of the joint program.
2023: Grateful to be part of the Microsoft Academic Partnership Grant.
2021: Started my PhD at Indian Institute of Technology Hyderabad.

Publications

A Unified View on Emotion Representation in Large Language Models [Recently accepted, yet to be published] author list: Maheswaran, A., Desarkar, M.S. In: 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2026)
Probing the Inherent Ability of Large Language Models for Generating Empathetic Responses Paper link, author list: Maheswaran, A., Chua, Caslon, Desarkar, M.S. In: 12th IEEE Swiss Conference on Data Science. SDS 2025
DAC: Quantized Optimal Transport Reward-based Reinforcement Learning Approach to Detoxify Query Auto-Completion Paper link Author list: Maheswaran, A., Maurya, K.K., Gupta, M., Desarkar, M.S, in SIGIR 2024
DQAC: Detoxifying Query Auto-Completion with Adapters Publication, Paper link, Author list: Maheswaran, A., Maurya, K.K. Gupta, M., Desarkar, M.S. In: Advances in Knowledge Discovery and Data Mining.PAKDD 2024
Adversarial Unsupervised Domain Adaptation for Hand Gesture Recognition Using Thermal Images Paper link, Author list: A. Dayal, M. Aishwarya, S. Abhilash, C. K. Mohan, A. Kumar and L. R. Cenkeramaddi, in IEEE Sensors Journal, vol. 23, no. 4, pp. 3493-3504, 15 Feb.15, 2023, doi: 10.1109/JSEN.2023.3235379