About

I’m a CS PhD Candidate in the ConvAI Lab at UIUC, advised by Prof. Dilek Hakkani-Tür and Prof. Gokhan Tur. Previously, I worked on social recommendation models at TikTok, personalized multimodal search at Verneek AI, and led speech recognition and synthesis research at Walmart India, contributing to a multilingual voice assistant. Academically, I hold a Master’s in Computer Science (Machine Learning) from Georgia Tech and a Bachelor’s in Computer Science with top honors from IIT Varanasi. For further details, please refer to my CV.

My current research interests are in reasoning in LLMs. I am interested in understanding which design attributes of “human-like reasoning” are missing in current LLMs, particularly abstraction and generalization, and how to measure and improve these behaviors in a reliable, interpretable way. In ML terms, I’m interested in efficient and adaptive abstractions, weak-to-strong and easy-to-hard generalization, and interpretability approaches that help us understand when and why reasoning succeeds or fails.

Apart from research and academics, I nurture my self with running (3x half), biking (50+ miles at a time), photography, reading, and crocheting.

Publications & Preprints

Aniket Anand=, Janvijay Singh=, Zhewei Sun, Dilek Hakkani-Tür, Nick Feamster. Measuring, Localizing, and Ablating Alignment Signatures in LLMs. arXiv preprint, May 2026.

Pardis Sadat Zahraei, Janvijay Singh, Gokhan Tur, Dilek Hakkani-Tür. Emergent Unfaithfulness: How Alignment Training Causes Language Models to Silently Override Task Faithfulness. Proceedings of the Workshop on Towards Knowledgeable Foundation Models, at ACL 2026, San Diego, United States, 2026.

Janvijay Singh, Dilek Hakkani-Tür. Do LLMs Encode Functional Importance of Reasoning Tokens?. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026, Main), San Diego, United States, July 2026.

Janvijay Singh, Austin Xu, Yilun Zhou, Yefan Zhou, Dilek Hakkani-Tür, Shafiq Joty. On the Shelf Life of Fine-Tuned LLM Judges: Future Proofing, Backward Compatibility, and Question Generalization. Proceedings of the 14th International Conference on Learning Representations (ICLR 2026), Rio de Janeiro, Brazil, April 2026.

Yefan Zhou, Austin Xu, Yilun Zhou, Janvijay Singh, Jiang Gui, Shafiq Joty. Variation in Verification: Understanding Verification Dynamics in Large Language Models. Proceedings of the 14th International Conference on Learning Representations (ICLR 2026), Rio de Janeiro, Brazil, April 2026.

Takyoung Kim=, Janvijay Singh=, Shuhaib Mehri=, Emre Can Acikgoz, Sagnik Mukherjee, Nimet Beyza Bozdag, Sumuk Shashidhar, Gokhan Tur, Dilek Hakkani-Tür. AURA: A Diagnostic Framework for Tracking User Satisfaction of Interactive Planning Agents. IEEE Transactions on Audio, Speech and Language Processing, 2026. Also presented at MTI-LLM at NeurIPS 2025, San Diego, Decemeber 2025.

Show more.

Janvijay Singh, Vilém Zouhar, Mrinmaya Sachan. Enhancing Textbooks with Visuals from the Web for Improved Learning. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP 2023, Main), Singapore, December 2023.

Janvijay Singh=, Mukund Rungta=, Diyi Yang, Saif Mohammad. Forgotten Knowledge: Examining the Citational Amnesia in NLP. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL 2023, Main), Toronto, Canada, July 2023. Best Paper Honourable Mention

Janvijay Singh, Fan Bai, Zhen Wang. Entity Tracking via Effective Use of Multi-Task Learning Model and Mention-guided Decoding. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2023, Main), Dubrovnik, Croatia, May 2023.

Mukund Rungta=, Janvijay Singh=, Saif Mohammad, Diyi Yang. Geographic Citation Gaps in NLP Research. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing (EMNLP 2022, Main), Abu Dhabi, UAE, December 2022.

Janvijay Singh, Anshul Wadhawan. Entity Recognition in Wet Lab Protocols using Structured Learning Ensemble and Contextualised Embeddings. Proceedings of the Sixth Workshop on Noisy User-generated Text (W-NUT 2020), at EMNLP 2020, Online, November 2020. Shared Task Winner

Janvijay Singh. Sentence and List Extraction in Noisy PDF Text Using a Hybrid Deep Learning and Rule-Based Approach. Proceedings of the Second Workshop on Financial Technology and Natural Language Processing (FinNLP 2020) at IJCAI-PRICAI 2020, Kyoto, Japan, January 2020. Shared Task Winner

Janvijay Singh, Raviraj Joshi. Background Sound Classification in Speech Audio Segments. Proceedings of the Tenth International Conference on Speech Technology and Human-Computer Dialogue (SpED 2019), Timișoara, Romania, October 2019.

Show less.

= denotes equal author contribution.

Teaching

I have served as a Teaching Assistant for the following courses:

• CS 546: Advanced Topics in NLP, Fall 2025, University of Illinois, Urbana-Champaign.
• CS 7643: Deep Learning, Spring 2023, Georgia Institute of Technology, Atlanta.
• CSE 6740: Computational Data Analysis, Fall 2022, Georgia Institute of Technology, Atlanta.
• CS 7650: Natural Language Processing, Spring 2022, Georgia Institute of Technology, Atlanta.
• CS 4641: Machine Learning, Fall 2021, Georgia Institute of Technology, Atlanta.
• CSO 101: Computer Programming and Linux, Fall & Spring 2017, Indian Institute of Technology, Varanasi.

Contact

The best way to reach me is via email: jvsingh2 [at] illinois [dot] edu OR janvijay [dot] singh [dot] cse14 [at] gmail [dot] com