# 几何金丝雀：基于表征稳定性预测可控性与检测漂移

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-04-20 08:00
- AIHOT 链接：https://aihot.virxact.com/items/cmo80g0e703boslml5qxc9v0v
- 原文链接：https://arxiv.org/abs/2604.17698

## AI 摘要

几何稳定性为语言模型部署提供双重诊断。监督式Shesha通过测量任务对齐的表征稳定性，在35-69个模型中以0.89-0.97相关系数精准预测线性可控性；无监督稳定性虽在可控性预测上失效（ρ≈0.10），却在漂移检测中表现优异：较CKA捕捉近2倍（Llama中5.23倍）几何变化，于73%模型中提前预警，假阳性率较Procrustes低6倍。两者分别适用于部署前可控性评估与部署后监控。

## 正文

Reliable deployment of language models requires two capabilities that appear distinct but share a common geometric foundation: predicting whether a model will accept targeted behavioral control, and detecting when its internal structure degrades. We show that geometric stability, the consistency of a representation's pairwise distance structure, addresses both. Supervised Shesha variants that measure task-aligned geometric stability predict linear steerability with near-perfect accuracy (ρ= 0.89-0.97) across 35-69 embedding models and three NLP tasks, capturing unique variance beyond class separability (partial ρ= 0.62-0.76). A critical dissociation emerges: unsupervised stability fails entirely for steering on real-world tasks (ρapprox 0.10), revealing that task alignment is essential for controllability prediction. However, unsupervised stability excels at drift detection, measuring nearly 2times greater geometric change than CKA during post-training alignment (up to 5.23times in Llama) while providing earlier warning in 73\% of models and maintaining a 6times lower false alarm rate than Procrustes. Together, supervised and unsupervised stability form complementary diagnostics for the LLM deployment lifecycle: one for pre-deployment controllability assessment, the other for post-deployment monitoring.