Sheer El-Showk

Also known as Sheer El Showk

Sheer El-Showk is a physicist who turned to AI research and is now a member of technical staff at Anthropic. He studied physics and mathematics at UC Berkeley and earned a doctorate at the University of Amsterdam in 2009. As a postdoctoral researcher at CEA Saclay and CERN, he co-authored the 2012 and 2014 papers that studied the 3D Ising model with the conformal bootstrap, and in 2015 he co-authored two bootstrap papers on supersymmetric theories. In 2022, he was among the authors of the papers on reinforcement learning from human feedback, on language-model self-knowledge and on Constitutional AI, and a core contributor to the training infrastructure behind Softmax Linear Units. His research at Anthropic covers ML safety, LLM evaluation and scalable oversight.

Now at: Anthropic
Works: 46
Citations: 4,963
h-index: 27
Badges: Bridge
Wayfarer

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

1996–2000

Physics and math at Berkeley

He studies physics and mathematics at the University of California, Berkeley, from 1996 to 2000 and earns a BA.
2009

Doctorate at Amsterdam

He earns a PhD at the University of Amsterdam in 2009, advised by Jan de Boer. His dissertation is titled "Puzzles in Quantum Gravity: What can Black Hole Microstates teach us about Quantum Gravity?"
2012

The 3D Ising bootstrap

As a postdoctoral researcher at CEA Saclay, he gives the talk "The Resurgent Bootstrap and the 3D Ising Model" at the KITP in April 2012. In 2012, he co-authors "Solving the 3D Ising Model with the Conformal Bootstrap" with Miguel F. Paulos, David Poland, Slava Rychkov, David Simmons-Duffin and Alessandro Vichi, which finds that the 3D Ising model lies at a corner point on the boundary of the allowed parameter space of 3D conformal field theories.
2014

Precise Ising critical exponents

In 2014, he co-authors the sequel "Solving the 3d Ising Model with the Conformal Bootstrap II" with Paulos, Poland, Rychkov, Simmons-Duffin and Vichi. The paper conjectures that the 3d Ising spectrum minimizes the central charge c among unitary solutions to crossing symmetry, and it finds a sharp transition in the operator spectrum at the Ising dimension Δσ = 0.518154(15).
2015

Supersymmetric bootstrap at CERN

In 2015, he co-authors two bootstrap papers on supersymmetric theories with Nikolay Bobev, Dalimil Mazac and Miguel F. Paulos. The first, on the three-dimensional supersymmetric Ising model, finds a kink at the expected location of the critical three-dimensional N=2 Wess-Zumino model. In March 2015, as a postdoctoral fellow at CERN, he speaks at an ICTP seminar on the second paper's topic, "Bootstrapping SCFTs with four supercharges".
2022

RLHF and Softmax Linear Units

The 2022 paper "Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback" lists him among its authors; it applies preference modeling and RLHF to finetune language models into helpful and harmless assistants. He is also a core contributor to the training infrastructure behind the Softmax Linear Units paper of June 2022.
2022

Self-knowledge and Constitutional AI

He is among the authors of "Language Models (Mostly) Know What They Know" in July 2022, which tests whether language models can evaluate the validity of their own claims and predict which questions they can answer correctly. In December 2022, he is among the authors of "Constitutional AI: Harmlessness from AI Feedback", which trains a harmless assistant through self-improvement with a list of rules or principles as the only human oversight.
As of October 2026

Member of technical staff

He is a member of technical staff at Anthropic, a role he has held since December 2021. His research covers ML safety, LLM evaluation and scalable oversight.