面向充分必要电路的全局信息阈值化
Global Information Thresholding for Sufficient and Necessary Circuits
为视觉模型（如 ViT、CLIP 视觉编码器）的电路级机制分析提供了一套无需人工设定电路规模的阈值选取流程：先用保号积分梯度给边打分，再按行为保留准则确定全局截断点，可直接迁移到视觉 Transformer 的注意力头/MLP 电路发现与充分性-必要性评估。
原文依据 We propose a procedure that scores edges with sign-preserving integrated gradients and then chooses a single global cutoff by a performance-retention criterion
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论机制可解释性旨在寻找既自身充分、又在对模型行为进行干预时必要的电路，然而许多电路发现流程仍依赖人工固定的电路预算。我们提出一种流程：先用保号积分梯度为边打分，再依据性能保留准则选取单一全局截断值以完成电路选择。这使得电路规模成为保留行为的后果，而非人工选定的预算。在机制可解释性基准上，这种经保留率校准的阈值化在多个任务与模型上得到的电路在 CPR/CMD 上具有竞争力。在我们针对 GPT-2 IOI 的概率空间诊断中，它相对 EAP-IG 基线提升了充分性，同时也改善了所报告的 Necessity 诊断指标。阈值消融与稳定性检验进一步表明，所选工作点在自充分性与面向必要性的诊断之间取得了最佳的整体平衡。
英文原文摘要 1011 字符
Mechanistic interpretability seeks circuits that are both sufficient on their own and necessary to a model's behavior under intervention, yet many circuit-discovery pipelines still rely on manually fixed circuit budgets. We propose a procedure that scores edges with sign-preserving integrated gradients and then chooses a single global cutoff by a performance-retention criterion for circuit selection. This makes circuit size a consequence of retained behavior rather than a manually selected budget. On the Mechanistic Interpretability Benchmark, this retention-calibrated thresholding yields circuits that are competitive on CPR/CMD across multiple tasks and models. On our GPT-2 IOI probability-space diagnostic, it improves sufficiency relative to the EAP-IG baseline while also improving the reported necessity diagnostics. Threshold ablations and stability checks further show that the selected operating point yields the best overall balance between self-sufficiency and necessity-oriented diagnostics.