发表机构
Institute for Gravitation and the Cosmos, Department of Physics, Pennsylvania State University; School of Physical Science and Engineering, Tongji University; State Key Laboratory of Autonomous Intelligent Unmanned Systems, MOE Frontiers ScienceCenter for Intelligent Autonomous Systems, Tongji University; School of Computer Science, Georgia Institute of Technology; School of Physics, Georgia Institute of Technology; Lawrence Berkeley National Laboratory（宾夕法尼亚州立大学引力与宇宙学研究所物理系; 同济大学物理科学与工程学院; 同济大学自主智能无人系统省部共建教育部前沿科学中心、自主智能无人系统国家重点实验室; 佐治亚理工学院计算机学院; 佐治亚理工学院物理学院; 劳伦斯伯克利国家实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出一种Transformer强化学习算法，融入已知对称性与关系，求解平面N=4超杨-米尔斯理论高圈散射振幅，降低先验需求，确保输出满足关系，通过赋值传播与MCTS搜索克服状态规模增长。
AI 中文摘要
我们提出了一种Transformer强化学习算法，用于学习求解平面N = 4超杨-米尔斯理论中的高圈级散射振幅。该算法通过将先前推导出的对称性和关系融入学习过程，改进了以往仅基于Transformer的结果。这大大降低了求解整个问题所需预先已知的解的比例。另一个优点是，该算法还确保每个输出都满足已知关系集。模型并非独立预测所有系数，而是提出赋值方案，并通过精确线性关系进行传播，当仅靠传播不足时，MCTS会在赋值方案上进行搜索。这对于机器学习方法向更高圈级推广至关重要，因为若无此机制，将无法克服状态规模阶乘增长的问题，也无法与其他方法得出的结果进行比较。利用形状因子的符号表示，我们将问题框架化为学习离散序列与整数系数之间的映射。
英文摘要
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.
