电路指纹：答案标记如何编码其几何路径
Circuit Fingerprints: How Answer Tokens Encode Their Geometrical Path
AI总结：
通过几何对齐实现电路发现，揭示transformer电路本质上是几何结构，实现可控引导并提升情感分类准确率。
摘要：
在transformer中，电路发现和激活引导发展为独立的研究线索，但两者都操作于同一表征空间。它们是同一底层结构的两种视角吗？我们展示了它们遵循单一几何原理：孤立处理的答案标记编码产生它们的方向。这种电路指纹假设使电路发现无需梯度或因果干预——通过纯粹的几何对齐恢复与梯度方法相当的结构。我们在四个模型家族的标准基准（IOI、SVA、MCQA）上验证了这一点，实现了与梯度方法相当的电路发现性能。识别电路组件的方向也使可控引导成为可能——在情感分类任务中达到69.8%的准确率，而指令提示仅为53.1%，同时保持事实准确性。除了方法开发外，这种读写二元性揭示了transformer电路本质上是几何结构：可解释性和可控性是同一对象的两个方面。
英文摘要：
Circuit discovery and activation steering in transformers have developed as separate research threads, yet both operate on the same representational space. Are they two views of the same underlying structure? We show they follow a single geometric principle: answer tokens, processed in isolation, encode the directions that would produce them. This Circuit Fingerprint hypothesis enables circuit discovery without gradients or causal intervention -- recovering comparable structure to gradient-based methods through geometric alignment alone. We validate this on standard benchmarks (IOI, SVA, MCQA) across four model families, achieving circuit discovery performance comparable to gradient-based methods. The same directions that identify circuit components also enable controlled steering -- achieving 69.8\% emotion classification accuracy versus 53.1\% for instruction prompting while preserving factual accuracy. Beyond method development, this read-write duality reveals that transformer circuits are fundamentally geometric structures: interpretability and controllability are two facets of the same object.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
