Learning Scattering Amplitudes with Transformer Reinforcement Learning
Abstract
We introduce a transformer reinforcement learning algorithm that learns to solve high loop-level scattering amplitudes in planar N = 4 Super Yang-Mills theory. Our algorithm improves on previous transformer-only results by incorporating previously derived symmetries and relationships into the learning algorithm. This results in a greatly decreased fraction of the solution that needs to be known a priori to solve the entire problem. An additional benefit is that our algorithm also ensures that every output obeys the set of known relationships. Rather than predicting all coefficients independently, the model proposes assignments that are propagated through exact linear relations, while MCTS searches over assignments when propagation alone is insufficient. This is crucial to the generalization of machine learning approaches to higher loops, as without this, there is no way to overcome the factorially scaling of state sizes and compare to results derived via other methods. Using the symbology representation of the form factor, we frame the problem as learning a mapping between discrete sequences and integer coefficients.
Related
- Atomic Design Transformer: Scaffold-Conditioned 3D Molecule Generation with xTB-Verified Reinforcement Learning
- Scattering amplitudes and resonant processes in QED with chiral chemical potential and chiral magnetic conductivity
- Scattering Equations as the lowest order K-identities in the calculation of Stringy Scaling of Hard String Scattering Amplitudes
- Scattering amplitudes from quantum hardware a la RESOs
- Light-by-light scattering at three loops in massless QCD and QED: amplitudes and cross sections
BibTeX
@misc{velie2026learning,
title = {{Learning Scattering Amplitudes with Transformer Reinforcement Learning}},
author = {Philip Velie and Tianji Cai and Piyush Jha and Vijay Ganesh and Aishik Ghosh},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.32096},
eprint = {2609.32096},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.32096},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Learning Scattering Amplitudes with Transformer Reinforcement Learning
AU - Philip Velie
AU - Tianji Cai
AU - Piyush Jha
AU - Vijay Ganesh
AU - Aishik Ghosh
PY - 2026
DA - 2026/09/26
PB - arXiv
DO - 10.48550/arxiv.2609.32096
UR - https://arxiv.org/abs/2609.32096
AB - We introduce a transformer reinforcement learning algorithm that learns to solve high loop-level scattering amplitudes in planar N = 4 Super Yang-Mills theory. Our algorithm improves on previous transformer-only results by incorporating previously derived symmetries and relationships into the learning algorithm. This results in a greatly decreased fraction of the solution that needs to be known a priori to solve the entire problem. An additional benefit is that our algorithm also ensures that every output obeys the set of known relationships. Rather than predicting all coefficients independently, the model proposes assignments that are propagated through exact linear relations, while MCTS searches over assignments when propagation alone is insufficient. This is crucial to the generalization of machine learning approaches to higher loops, as without this, there is no way to overcome the factorially scaling of state sizes and compare to results derived via other methods. Using the symbology representation of the form factor, we frame the problem as learning a mapping between discrete sequences and integer coefficients.
ER -