Related Professors at NYU

A collection of related professors. Inclusion does not imply an affiliation with the association.

• Pavel Izmailov. Reinforcement learning, reasoning, and alignment . Researcher at Anthropic, formerly OpenAI superalignment.
• Greg Durrett. NLP and machine learning. Many papers on alignment and interpretability.
• Qi Lei. Weak-to-strong generalization, distribution shift, and inference-time safeguards . Google DeepMind faculty.
• He He. Reward hacking , control , and monitoring . (On sabbatical fall '27)
• Sam Bowman. Alignment. Anthropic. (Long-term leave)
• Sydney Levine. Psychology professor working at "the intersection of cognitive science of human moral judgement and AI safety". Visiting scientist at Google Deepmind. Recent papers in alignment, value prioritization evaluation, and meta cognition.
• Sebastian Benthall. Research director, Information Law Institute. Work on casual decision processes and societal lock-in .
• Nadav Brandes. Biosecurity risks from genomic AI models.
• Gustavo Sandoval. Work on interpretability and secure code generation .

• Jonathan Mann. Adjunct professor and Good Judgement superforecaster. Writes about forecasting and AI safety on Substack.
• Kyunghyun Cho. Mainly focuses on AI for biology and medicine, but has published on alignment, unlearning, and interpretability.
• Joan Bruna. Mostly deep learning theory. One interpretability paper.
• Eunsol Choi. Partial focus on continual learning. Papers on LLM behavioral evaluation and implicit user feedback for training.
• Arthur Jacot. Deep learning theory. Explores which notion of functional complexity the algorithms/circuits of DNNs implicitly minimize which could help ground mechanistic interpretability circuit research.
• Mengye Ren. Mostly representation and continual learning. Papers on scaleable oversight and forecast calibration.
• Julia Kempe. Some work on alignment, mechanistic interpretability, and hallucination detection. Director of research at AmiLabs.
• Chinmay Hegde . Red-team jailbreaking and unlearning .
• Siddharth Garg . Red-team jailbreaking and interpretability of insecure code generation.
• Andrew Gordon Wilson. Mostly theory of deep learning, but has touched out-of-distribution detection and hallucination detection.
• Rajesh Ranganath. Quantifying tail-risk of LLM outputs and explanation faithfulness.
• Ramesh Karri. Security and trust for hardware, lately encompassing LLMs. Recent papers on jailbreak benchmarking and unlearning
• Gautam Kamath. Mostly differential privacy, but has published on data poisoning and unlearning.
• Tal Linzen. Interpretability[22] and introspection.
• Haneul Yoo. Natural language processing, mostly focusing on non-English languages. One paper on jailbreaking LLMs by switching languages .
• Florentin Guth. Models the weight distributions of trained networks to describe what they encode . Could help formalize the notion of features in mechanistic interpretability.