Counterfactual Predictions in Scientific Emulators Without Controlled Experiments

在无需受控实验的情况下实现科学模拟器的反事实预测

Abstract: Many scientific questions require reasoning about what was never observed: What if the conditions, interventions, or history had been different? Models can predict accurately on observed data yet fail on such what-if queries when correlated inputs are varied independently. A common remedy is to add controlled simulation data in which these factors are explicitly disentangled, but this requires access to a simulator, can be computationally expensive, and inherits the simulator’s modeling assumptions.

摘要： 许多科学问题需要对从未观察到的情况进行推理：如果条件、干预措施或历史情况有所不同，会发生什么？模型虽然能在观测数据上进行准确预测，但在处理这类“如果……会怎样”的查询时，一旦相关输入被独立改变，模型往往会失效。一种常见的补救措施是添加受控模拟数据，将这些因素显式解耦，但这需要访问模拟器，计算成本可能很高，且会继承模拟器的建模假设。

We introduce ReRoute, a framework for targeted scientific what-if prediction that combines factual data with partial mechanistic knowledge, without requiring controlled intervention data for adaptation. ReRoute fixes the queried input of a pretrained backbone to a reference value, reintroduces its variation through a known mechanistic pathway, and fine-tunes on the original factual data, while leaving downstream effects to the learned dynamics.

我们引入了 ReRoute，这是一个针对科学“如果……会怎样”预测的框架。它将事实数据与部分机制知识相结合，无需受控干预数据即可进行适配。ReRoute 将预训练主干网络的查询输入固定为参考值，通过已知的机制路径重新引入其变化，并在原始事实数据上进行微调，同时将下游效应留给学习到的动力学模型处理。

We provide a causal identification result for this construction under explicit structural assumptions, with the core argument machine-checked in Lean. After showing that ReRoute achieves highly accurate counterfactual predictions in a controlled advection-diffusion system where exact responses are available, we turn to state-of-the-art climate emulation.

我们在明确的结构假设下为该架构提供了因果识别结果，其核心论点已通过 Lean 证明助手进行了机器验证。在证明了 ReRoute 在拥有精确响应的受控平流扩散系统中能实现高度准确的反事实预测后，我们将研究转向了最先进的气候模拟。

On held-out coupled-climate interventions, ReRoute reduces aggregate climate error by 18.2-31.8% under severe CO$_2$ distribution shifts while preserving skill under standard conditions, at a small fraction of the cost of retraining on additional controlled simulations, without even accounting for the substantial expense of generating such data.

在留出的耦合气候干预测试中，ReRoute 在严重的 CO$_2$ 分布偏移下将总气候误差降低了 18.2-31.8%，同时在标准条件下保持了预测能力。其成本仅为在额外受控模拟上进行重训练的一小部分，且这还未计入生成此类数据所需的大量开销。

Finally, on an emulator trained from historical ERA5 reanalysis, where no counterfactual reference exists, ReRoute preserves substantially more of the surface warming implied by the observed boundary conditions under a fixed-CO$_2$ counterfactual.

最后，在一个基于历史 ERA5 再分析数据训练的模拟器上（此处不存在反事实参考），ReRoute 在固定 CO$_2$ 的反事实条件下，保留了更多由观测边界条件所隐含的地表变暖趋势。