通过选择关键主成分，让语言模型探针在分布外欺骗检测中实现跨领域高精度识别。
Probe Generalization as Subspace Selection for OOD Deception Detection
- 从训练数据激活中选取少量主成分，提升探针对分布外样本的泛化能力。
- 在内幕交易报告和沙袋测试数据集上，性能差距缩小78%和25%。
- 用大模型判断主成分是否具可迁移性，适合关注模型鲁棒性的研究者。
线性探针可用于检测语言模型激活中的行为与概念，但在分布外样本上可能失效。在对Llama-3.1-8B-Instruct探针在三个保留的欺骗检测数据集上的泛化性能研究中发现，将输入投影到训练分布激活的少数主成分（PCs）上，即可实现接近直接在测试分布上训练探针的跨域迁移性能。进一步发现，可通过大模型裁判评分每个主成分的可迁移性：依据其最激活/最少激活样本是否暗示可迁移的欺骗方向，选取得分最高的主成分进行探测，使内幕交易报告任务的基线到理想性能差距缩小78%，沙袋测试任务缩小25%。源探针权重高的方向往往编码源域特有表面特征，而真正可迁移的方向则以更抽象的方式编码相同对比，且自然语言描述可捕捉其含义。总体表明，探针的分布外鲁棒性主要取决于子空间选择。
原文摘要 · Abstract (English)
Linear probes can be used to detect behaviors and concepts inside language model activations, but may fail to transfer to out-of-distribution examples. When studying the generalization performance of Llama-3.1-8B-Instruct probes over 3 held-out deception detection datasets, we find that projecting inputs onto a small subset of principal components (PCs) from the training distribution of activations enables cross-domain transfer that nearly matches the performance of probes trained directly on the test distribution. Furthermore, we find that PC interpretations can be used to find a subset of those transferable PCs. By using an LLM judge to score each PC on whether its most/ least activating examples imply a transferable deception direction, then probing on the highest-scoring PCs, we close the baseline-to-oracle gap by 78% on Insider Trading Report and by 25% on Sandbagging. The directions a source probe weights heavily appear to encode source-specific surface features, while the directions that actually transfer appear to encode the same contrast more abstractly, in a way natural language descriptions can capture. Broadly, our results suggest that the OOD robustness of probes is largely determined by subspace selection.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。