arXiv cs.AI· Man Liang, Xinzhao Cheng, Faizan Wajid·· 14 小时前AI 评分31
冻结 LLM 几何约束审计：可解码但不等于可行动
Decodable but Not Actionable: Auditing the Representation-Behavior Gap in Frozen LLMs for Geometric Constraints
AI 导读
研究团队提出基于参数化 CAD 的受控审计框架，在八个预训练模型及匹配的随机初始化对照上区分编码、表达与激活层控制的失败。预训练对局部关系的线性解码贡献远大于全局约束状态，而强局部可解码性并未转化为零样本预测能力，仍远低于监督探针。仅用 400 对标注关系做低秩微调即可显著提升三个代表性模型的生成表现，说明表征—行为差距可通过适配缩小。
正文 · 原文
Abstract:Large language models (LLMs) perform well on structured reasoning tasks, but what they encode and whether encoded information guides their behavior remain unclear. We introduce a controlled auditing framework that uses parametric CAD to distinguish failures of encoding, expression, and activation-level control. Applied to eight pretrained models and matched randomly initialized controls, our framework reveals that pretraining contributes substantially more to linear decoding of local relations than to global constraint status, which is already highly decodable from random representations. Yet strong local decodability does not translate into the models' zero-shot predictions, which remain far below supervised probe performance. Restoring clean internal activations can repair probe predictions, whereas simple activation steering does not provide reliable targeted control. With only 400 labeled relation pairs, low-rank fine-tuning substantially improves generation on three representative models, showing that the representation--behavior gap can be reduced through adaptation. Together, these findings show why decodability alone is insufficient evidence that encoded information will be expressed or controllable in model behavior.
| Comments: | Revised manuscript with expanded experiments, analyses, and discussion |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.17843 [cs.CL] |
| (or arXiv:2608.17843v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2608.17843 arXiv-issued DOI via DataCite |
Submission history
From: Man Liang [view email]
[v1]
Tue, 18 Aug 2026 14:40:28 UTC (1,567 KB)
[v2]
Tue, 27 Oct 2026 23:04:55 UTC (1,024 KB)
来源：arXiv cs.AI · arxiv.org