Manya Wadhwa

I am a PhD student at NYU (as a part of the TAUR lab), co-advised by Professor Greg Durrett and Professor Jessy Li. I am currently a Visiting Research Scientist at Meta/FAIR in NYC, working with Jesse Dodge. I am currently supported by the AIM Fellowship (AI Fellowship by Meta).

My interests lie in evaluating language models (CREATE, GENIE, EvalAgent, SmoothScore) and in finding ways to train and improve them (Detect-Critique-Refine, SkillFactory, Set-level RL). My focus has primarily been on non-verifiable tasks like short-form QA, long-form writing, scientific ideation, and creative problem solving.

I started my PhD at UT Austin, where I was a part of the UT NLP group.

Experience

During my PhD, I have worked with Adobe (Document Intelligence team) and Salesforce Research (Interactive AI team).

Before my PhD, I was an Associate Research Engineer with the R&D team at Goldman Sachs for three years. This team was led by Vijay Saraswat. I closely worked with Johannes Hoffart and Luciano Del Corro on problems in the space of Document AI and Conversation Understanding specific to the field of Finance.

I completed my Master's in Computer Science from Johns Hopkins University where I worked with Professor Mark Dredze. I did my undergrad from IIIT-Delhi where I majored in Computer Science.

News

• Sept 2026Started as a Visiting Research Scientist at Meta/FAIR in NYC!
• May 2026Interning with Salesforce Research in CA!
• April 2026Co-presenting SkillFactory at ICLR 2026! See you in Brazil!
• Feb 2026I am grateful to have also been selected for the Bloomberg Data Science PhD Fellowship (could not accept due to a prior commitment).
• Aug 2025Moved to NYU as a fourth year PhD student with my advisor!
• July 2025Cleared my qualification presentation at UT Austin!

Publications