arXiv cs.CL· Marcus Armstrong·· 11 小时前AI 评分32
语言模型对同一实体的首次回答如何影响后续回答
Written Once, Read Differently How Language Models Own Answers Steer Their Next Ones
AI 导读
研究在四个家族共 22 个开放权重模型（0.5B–72B 参数）上考察"自我承诺"现象：模型对某实体的首个是/否答案会写入残差流中的承诺方向，两个问题方向的相似度可预测其在留出实体上的耦合（ρ = 0.43–0.85）。
正文 · 原文
Abstract:When a language model answers a yes/no question, that answer changes how it answers a different question about the same entity. We study this \emph{self-commitment} in 22 open-weight models from four families (0.5B--72B parameters), fixing the model's first answer and measuring the shift in its second across 3,600 question pairs about 120 entities. Each answer writes a \emph{commitment direction} into the residual stream. The similarity of two questions' directions predicts their coupling on held-out entities ($\rho = 0.43$--$0.85$), and ablating a single direction shared across topics leaves at most 17\% of the effect in every model. How this direction is read is far less stable: the reading circuit spreads across more attention heads as models grow, instruction tuning makes more questions read it in the opposite sense, and in one model changing only the system prompt reverses the circuit's sign. Ablating the direction while sparing the attention sink removes most of the coupling at under 1.3\% perplexity cost in 18 of 19 instruction-tuned models, though it shifts standalone answers in two. Without the sink spared, the direction's alignment with the leading principal component of ordinary text predicts which models pay a high cost. Written once, the commitment is read differently across scale, training and prompt.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.03198 [cs.CL] |
| (or arXiv:2611.03198v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.03198 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Marcus Armstrong [view email]
[v1]
Mon, 2 Nov 2026 14:25:14 UTC (259 KB)
来源：arXiv cs.CL · arxiv.org