Steering Methods for Inference-Time Political Debiasing of LLMs
Anshika Agarwal
Stanford University
About
I graduated from Stanford with my bachelor's and master's degrees in Computer Science, concentrated in AI, with Honors and Distinction. I am currently at Meta Superintelligence Labs, working on applied research. Last summer, I was also at MSL building speech model infrastructure, with prior experience in applied AI at AWS and Quizlet. Understanding the societal impacts of AI is deeply important to me; I spent the spring of 2024 at the Oxford Internet Institute writing about digital economics and tech regulation. My research interests broadly span human-centered AI and sustainable computing.
Selected Works
Recipient of Stanford's Eric Roberts Prize for Best Paper on Ethics and Technology (2025)
The Translation Turing Test: Evaluating LLM-Mediated Translation for Cross-Lingual Interpersonal Communication
Advised by Dr. Jiaxin Pei and Dr. Diyi Yang
Written during my time abroad at Oxford, advised by Dr. Liam Bekrisky