Tag
This paper proposes a unified path-integral framework for generative modeling, showing that flow-based, diffusion-based, variational, and adversarial models arise from different evaluation principles of a single master action. It derives a one-loop correction that reduces tree-level error dramatically and introduces a response-weighted score-matching objective.
The author shares his reflections on Stephen Wolfram's A New Kind of Science, arguing that continuous calculus is only an engineering compromise made for a discrete universe in an era of scarce computing power, and discusses how computational irreducibility challenges the traditional paradigm of prediction.
New theoretical calculations resolve a 25-year muon g-2 puzzle, but create a clash with older experimental results, suggesting possible new physics.
A trio of Norwegian physicists theoretically investigated what happens when a mirror is suddenly removed while a single photon is in the process of reflecting, finding that the photon can split into multiple photons due to the nonlinearity of the event.
Leading theoretical physicist Yuji Tachikawa reports that Claude Fable solved a research problem that had stumped him and his collaborators for six months.
François Englert, the Belgian theoretical physicist who co-discovered the Higgs mechanism, has passed away at age 93. His work, along with Robert Brout and Peter Higgs, led to the discovery of the Higgs boson in 2012 and the Nobel Prize in Physics in 2013.
Physicists have identified a quantum property called 'magic' that may explain how matter tells space-time to curve, addressing a key challenge in holographic theories of quantum gravity.
A tweet observes AI agents collaboratively solving a difficult theoretical physics problem, demonstrating self-correction and equation derivation.
Hugging Face releases 'physics-intern', an agentic framework for theoretical physics research that doubles the performance of Gemini models on the CritPt benchmark and sets a new state-of-the-art compared to GPT-5.5 Pro.
Physics-intern is an agentic framework for theoretical physics that improves Gemini 3.1 Pro's performance on the CritPt benchmark from 17.7% to 31.4%, achieving a new state-of-the-art.
This paper investigates using Large Language Models, specifically Claude, interfaced with a Computer Algebra System (Maple) to perform algorithmic computations in theoretical physics, such as analyzing cosmological perturbations.
This paper introduces SCALAR, a structured critic-actor loop framework, to evaluate how different interaction patterns between AI agents improve reasoning in theoretical physics problems.
OpenAI researchers and collaborators have published a preprint showing that single-minus graviton tree amplitudes are nonzero under special kinematic conditions (the half-collinear regime), challenging long-held assumptions. Notably, GPT-5.2 Pro was used to derive the gravitational extension from prior gluon results, representing a significant AI-assisted breakthrough in theoretical physics.
GPT-5.2 assisted in deriving a new theoretical physics result showing that single-minus gluon tree amplitudes can be nonzero under specific half-collinear momentum conditions, challenging decades of assumptions in particle physics. The AI model identified patterns in complex Feynman diagram expressions and conjectured a general formula that was subsequently verified through formal proofs.