Arjun Panickssery
Bio
Arjun Panickssery is an AI safety researcher and entrepreneur based in the San Francisco Bay Area. He studied at the University of Illinois at Urbana-Champaign and has been active in AI alignment research through multiple programs and institutions. He participated in MATS (Machine Learning Alignment Theory Scholars), including an extension phase in London, where his research on the safety implications of LLM self-recognition produced the widely-cited paper "LLM Evaluators Recognize and Favor Their Own Generations" (co-authored with Samuel R. Bowman and Shi Feng), which demonstrated that frontier models such as GPT-4 can recognize their own outputs and exhibit self-preference bias that could undermine safety techniques like reward modeling and constitutional AI. He subsequently worked on scalable oversight benchmarks as part of MATS Summer 2024 and previously held roles at METR Evals and an AI risks organization. He is also building Zembla, an AI-powered platform for accelerated, individualized learning, and writes frequently about AI tutoring and education research.
Links
- Personal Website
- https://arjunpanickssery.com/
- Twitter / X
- LessWrong
- arjun-panickssery
Grants
from Long-Term Future Fund
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Mar 22, 2026, 2:27 PM UTC
- Created
- Mar 20, 2026, 2:47 AM UTC