Sima Noorani

Ph.D. student at the University of Pennsylvania

Advised by Hamed Hassani and George J. Pappas

I am a Ph.D. student at the University of Pennsylvania. I am also currently a Student Researcher at Google Research. My research focuses on developing rigorous statistical methodologies for reliable human-AI collaboration and trustworthy AI agents. In particular, I explore how to rethink uncertainty quantification to provide formal guarantees when humans and AI systems reason and make decisions together. Before beginning my Ph.D., I studied electrical engineering with a minor in computer science at Drexel University and interned at Lockheed Martin, Comcast, and Bristol Myers Squibb.

News & Talks

Sep 2026 “Strategic Decision Support for AI Agents” accepted to NeurIPS 2026.
Jul 2026 “Human–AI Collaborative Uncertainty Quantification” selected as a spotlight at the Statistical Frameworks for Uncertainty in Agentic Systems Workshop at ICML 2026.
Jun 2026 Started as a Student Researcher at Google Research.
May 2026 “When to Trust the Cheap Check: Weak and Strong Verification for Reasoning” accepted to ICML 2026 as a spotlight.
Nov 2025 Here are my slides on "Uncertainty Quantification for Generative & Collaborative AI" presented at Max Planck Institute for Software Systems. slides
Nov 2025 Here are my slides on “Human–AI Collaborative Uncertainty Quantification” presented at the Berkeley ANSR Reading Group. slides
Oct 2025 “Conformal Prediction Beyond the Seen” accepted to NeurIPS 2025. SlidesLive

Papers

Disagreement-Guided Weak-to-Strong Routing

S. Noorani*, S. Kiyani*, G. J. Pappas, H. Hassani

arXiv preprint, 2026.

[pdf]

Strategic Decision Support for AI Agents

S. Kiyani*, S. Noorani*, G. J. Pappas, H. Hassani

NeurIPS 2026.

[pdf]

Human–AI Collaborative Uncertainty Quantification

S. Noorani*, S. Kiyani*, G. J. Pappas, H. Hassani

ICML, 2026.

Spotlight at the Statistical Frameworks for Uncertainty in Agentic Systems Workshop at ICML 2026.

[pdf]

Multi-Round Human–AI Collaboration with User-Specified Requirements

S. Noorani*, S. Kiyani*, H. Hassani, G. J. Pappas

ICML, 2026.

[pdf]

When to Trust the Cheap Check: Weak and Strong Verification for Reasoning

S. Kiyani, S. Noorani, G. J. Pappas, H. Hassani

ICML 2026 (Spotlight).

Conformal Prediction Beyond the Seen: A Missing Mass Perspective for Uncertainty Quantification in Generative Models

S. Noorani*, S. Kiyani*, G. J. Pappas, H. Hassani

NeurIPS 2025.

Also accepted to the R2-FM Workshop at ICML 2025.

[pdf]

Conformal Risk Minimization with Variance Reduction

S. Noorani, O. Romero, N. Dal Fabbro, H. Hassani, G. J. Pappas

UAI 2026.

Also accepted to the R2-FM Workshop at ICML 2025.

[pdf]