Ameen Patel | Irvington HS, Stanford University

About Me

I am a senior in Irvington High School. I have done sycophancy research in the past, and currently, I work as a student researcher at Stanford University’s STAIR lab. I would like to thank Professor Sanmi Koyejo, Sang Truong, and Nathan Hu for mentoring and teaching me so much.

Research Interests

AI measurement: item response theory, safety
Mechanistic Interpretability: circuit tracing, automation, attribution graphs
Sycophancy: LLM behavior, emotional framing

News

[July 2026] Our paper about an IRT model for decomposing multilingual safety is accepted to COLM 2026.
[June 2026] Our paper about LLMs annotating attribution graphs is accepted to ICML Mechanistic Interpretability 2026.
[Oct. 2025] Our paper about emotional prompting with LLMs is accepted to AACL SRW 2025. Cited by Amazon researchers.

Publications

Why Do Safety Guardrails Degrade Across Languages?

Max Zhang^*, Ameen Patel^*, Sang Truong^†, Sanmi Koyejo^†
^* Equal Contribution
^† Equal Senior Authorship

Conference on Language Modeling (COLM), 2026.

Paper Code Poster Presentation

LLMs can Annotate Attribution Graphs

Ameen Patel^*, Max Zhang^*, Nathan Hu
^* Equal Contribution

International Conference on Machine Learning Mechanistic Interpretability Workshop (ICML), 2026.

Paper Code Poster Presentation

The Role of Emotional Stimuli and Intensity in Shaping Large Language Model Behavior

Ameen Patel^*, Felix Lee^*, Kyle Liang^*, Joseph Thomas
^* Equal Contribution

Asia-Pacific Chapter of the Association for Computational Linguistics Student Research Workshop (AACL SRW), 2025.

Paper Code Poster Presentation