How does the internal computation of a machine learning model transform
inputs into predictions? In this paper, we introduce a task called component
modeling that aims to address this question. The goal of component modeling is
to decompose an ML model's prediction in terms of its components -- simple
functions (e.g., convolution filters, attention heads) that are the "building
blocks" of model computation. We focus on a special case of this task,
component attribution, where the goal is to estimate the counterfactual impact
of individual components on a given prediction. We then present COAR, a
scalable algorithm for estimating component attributions; we demonstrate its
effectiveness across models, datasets, and modalities. Finally, we show that
component attributions estimated with COAR directly enable model editing across
five tasks, namely: fixing model errors, ``forgetting'' specific classes,
boosting subpopulation robustness, localizing backdoor attacks, and improving
robustness to typographic attacks. We provide code for COAR at
this https URL .

这篇论文介绍了一个名为 component modeling 的任务，旨在解决机器学习模型如何将输入转化为预测的内部计算方式，并提出了 COAR 算法用于估计组件归因，证明其在模型编辑方面的有效性。

通过建模模型计算对预测结果进行分解和编辑

Decomposing and Editing Predictions by Modeling Model Computation

A fundamental task for propositional logic is to compute models of
propositional formulas. Programs developed for this task are called
satisfiability solvers. We show that transition systems introduced by
Nieuwenhuis, Oliveras, and Tinelli to model and analyze satisfiability solvers
can be adapted for solvers developed for two other propositional formalisms:
logic programming under the answer-set semantics, and the logic PC(ID). We show
that in each case the task of computing models can be seen as "satisfiability
modulo answer-set programming," where the goal is to find a model of a theory
that also is an answer set of a certain program. The unifying perspective we
develop shows, in particular, that solvers CLASP and MINISATID are closely
related despite being developed for different formalisms, one for answer-set
programming and the latter for the logic PC(ID).

该研究将过渡系统用于计算命题公式的模型，证明了其适用于逻辑规划和 PC（ID）逻辑中，通过计算答案集表示的模型，提供了一种统一的视角，为不同命题形式开发的解决程序提供联系。