发表机构
NVIDIA; Seamless Labs; Oak Ridge National Laboratory（英伟达; 无缝实验室; 橡树岭国家实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究在Lattice和Hohlraum基准上基准测试两种神经代理架构（Transolver和BSMS-MGN）用于线性辐射输运模拟，发现归纳偏置偏好依赖架构，下游效用需按QoI评估，并公开训练配方与数据。
AI 中文摘要
线性辐射输运方程（RTE）构成核工程、惯性约束聚变、医学成像和天体物理学中设计分析任务模拟的基础，但以工程保真度解析高维相空间仍然昂贵，以至于设计优化、不确定性量化和参数扫描等外环工作流在传统求解器上经常受预算限制。神经代理模型有望通过将模拟成本分摊到数千次下游查询来缓解这一瓶颈，但使代理模型在某一输运问题上准确的架构选择和工程化归纳偏置并不能直接跨模型族迁移。我们在规范Lattice和Hohlraum基准上，将两种参数匹配的神经代理架构——物理注意力Transolver和多尺度图网络Bi-Stride Multi-Scale MeshGraphNet（BSMS-MGN）——作为二维线性RTE最终时刻粒子浓度的端到端近似进行基准测试。对傅里叶特征和区域加权训练损失的消融研究揭示了强烈依赖架构的归纳偏置偏好，表明物理信息代理工作流中常见的设计选择必须针对每种架构重新审视，而非跨模型族直接导入，且下游效用取决于每个感兴趣量（QoI）的敏感性，而非单一的场级评分。模型训练配方、训练数据和评估流程随论文一同发布，以支持复现、迁移到相关输运问题，以及在更大外环模拟工作流中作为摊销前向模型组件的评估。
英文摘要
Linear radiation transport equations (RTEs) form the simulation foundations underpinning design and analysis tasks in nuclear engineering, inertial confinement fusion, medical imaging, and astrophysics, but resolving the high-dimensional phase space at engineering fidelity remains expensive enough that outer-loop workflows, such as design optimization, uncertainty quantification, and parameter sweeps, are routinely budget-bound on traditional solvers. Neural surrogates promise to relax this bottleneck by amortizing simulation cost across thousands of downstream queries, but the architectural choices and engineered inductive biases that make a surrogate accurate on one transport problem do not transfer straightforwardly across model families. We benchmark two parameter-matched neural surrogate architectures, the physics-attention Transolver and the multi-scale graph network Bi-Stride Multi-Scale MeshGraphNet (BSMS-MGN), as end-to-end approximations of the final-time particle concentration for the two-dimensional linear RTE on the canonical Lattice and Hohlraum benchmarks. An ablation across Fourier features and region-weighted training loss exposes strongly architecture-dependent inductive-bias preferences, indicating that design choices common to physics-informed surrogate workflows must be revisited per architecture rather than imported across model families, and that downstream utility depends on per-QoI sensitivity rather than a single field-level score. The model training recipe, training data, and evaluation pipeline are released alongside this paper to support reproduction, transfer to related transport problems, and evaluation as amortized forward-model components in larger outer-loop simulation workflows.
