PSI 方法:无需反事实标注即可学习可解释的因果对齐子空间
Programmable Semantic Interventions: Learning Causal Carriers from Behavioral Programs
搞可解释性的人可以看看这篇:以前做因果对齐得先写好算法再人工标注反事实,现在只需声明一个语义程序,答案直接从冻结模型里读。MIB 基准上效果和有监督对齐打平,还能顺手做无 donor 引导,定位加控制命中 94.8%。
arXiv 论文提出 Programmable Semantic Interventions(PSI),一种从行为程序中学习因果载体的方法。传统因果抽象需要对每对提示提供反事实标签,而 PSI 让研究者直接定义语义程序:变量可产生哪些答案、干预应产生哪一个。
技术上,PSI 学习一个低秩激活子空间,交换该子空间即可按程序声明改变模型输出,预期答案直接从冻结模型自身读取,因此无需整理好的配对数据和真实标签。在 MIB 基准上,PSI 在相同数据和训练预算下与有监督对齐性能持平。
应用方面,语义程序的声明还固定了一个语义已知的坐标系,实现无 donor 的模型引导(steering):联合训练定位与控制后,PSI 在held-out 请求上实现 investigator 所选答案的 94.8%,无需 donor 提示。对于保留 gold 标签的自然科学问题,PSI 比用 gold 标签训练的子空间更贴合模型自身答案。
Programmable Semantic Interventions: Learning Causal Carriers from Behavioral Programs
Causal abstraction explains a frozen network by aligning its activations with the variables of a simpler, interpretable model, and checking that interventions on the two agree. Finding such an alignment currently requires a counterfactual dataset: for every pair of prompts, the answer the network ought to give once the variable has been changed. In practice, those labels are produced by an executable algorithm for the behavior, written before studying the model supposed to implement it. However, the criterion these methods optimize only leverages the answer an intervention should produce. Here, we develop an approach that lets the investigator define that relationship directly: the answers the variable can produce, and which of them an intervention should yield. We call this interface a semantic program and introduce Programmable Semantic Interventions (PSI): a low-rank activation subspace learned so that interchanging it changes the model's answer as the program declares, with the expected answers read from the frozen model itself. On the MIB benchmark PSI performs on par with supervised alignment at matched data and training budget, without curated pairs, and it needs no ground-truth labels; while on natural science questions with gold labels withheld, it tracks the model's own answers better than a subspace trained on gold labels. The declaration also fixes a coordinate system with known semantics, enabling donor-free steering: jointly training PSI for both localization and control realizes 94.8% of held-out requests for answers the investigator chooses, without the need for donor prompts.
This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include interpretability, PSI, MIB, steering.
Source attribution: curated from arXiv cs.AI (@Paolo Muratore, Mackenzie Weygandt Mathis).