Prasann Singhal
प्रसन्न सिंघल

Undergraduate Student, Computer Science / Linguistics
The University of Texas at Austin

prasanns [at] cs.utexas.edu   PrasannS   prasann_singhal

I am a fourth year undergraduate student at UT Austin, advised by Greg Durrett, where I am a part of the TAUR lab (Text Analysis, Understanding, and Reasoning). I'm also affiliated as a part-time researcher with some cool folks at Scale AI.

My research focuses on Natural Language Processing and topics in natural language generation and model interpretability. I'm currently working on getting a better understanding for how properties of data affect what gets learned by different kinds of learning algorithms for generation. I also have done some work in understanding and improving RLHF algorithms (particularly focused on reward modeling and things like online vs offline supervision), efficiency in decoding algorithms, and using explanations to predict large language model out-of-distribution robustness.

Prior to research, I've also had the chance to participate in nearly 20 hackathons, and I've made a lot of fun and interesting projects.

Feel free to send me an email if you'd like to chat about data, RLHF, text generation, interpretability, or NLP in general! Likewise, if you're a high-schooler or undergrad who's interested in getting into machine learning / research, I'm always happy to give advice if you feel like I can be of any help.

Publications

To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning

Zayne Sprague, Fangcong Yin, Juan Diego Rodriguez, Dongwei Jiang, Manya Wadhwa, Prasann Singhal, Xinyu Zhao, Xi Ye, Kyle Mahowald, and Greg Durrett. arXiv 2024.

A Long Way to Go: Investigating Length Correlations in RLHF code

Prasann Singhal, Tanya Goyal, Jiacheng Xu, and Greg Durrett. (Oral Spotlight) Proceedings of COLM 2024.

D2PO: Discriminator-Guided DPO with Response Evaluation Models code

Prasann Singhal, Nathan Lambert, Scott Niekum, Tanya Goyal, and Greg Durrett. Proceedings of COLM 2024.

EEL: Efficiently Encoding Lattices for Reranking code

Prasann Singhal, Jiacheng Xu, Xi Ye, and Greg Durrett. Proceedings of ACL 2023.

Assessing Out-of-Domain Language Model Performance from Few Examples

Prasann Singhal*, Jarad Forristal*, Xi Ye, and Greg Durrett. Proceedings of EACL 2023.

Invited Talks

[Jul. 2024] A Long Way to Go in RLHF @ SEAL Reading Group, Scale AI

[Feb. 2024] A Long Way to Go in RLHF @ UT Austin LIN 393 Seminar

[Nov. 2023] A Long Way to Go in RLHF @ IST & Unbabel Seminar

Experience

Scale AI Research Intern / Part-time Researcher. Summer 2024-Present.

UT Austin TAUR Lab Undergraduate Research Assistant, Natural Language Processing. Fall 2021-Present.

Drishya.ai. ML Intern. Summer 2020 [Led project for simulating and multi-agent optimization of renewable microgrids].

Service/Teaching

CS 388 (NLP) TA (Fall 2024)

UT Austin Directed Reading Program Mentor (Spring 2024)

Reviewer: EMNLP (22)

Founder / Teacher - Katy HACK Initiative: I spent 3 years starting/running CS education programs in local elementary / junior high schools

Volunteer Teacher - I spent a summer teaching English, Computer Fundamentals in a rural village in Gujarat