逐字解析越南法律标题，揭示嵌入向量的动态演化规律
Reading a Legal Question Word by Word: Embedding Trajectories of 2,144 Vietnamese Legal Headlines
- 通过逐词编码法律标题，追踪嵌入向量在检索过程中的动态变化
- 多数标题在6-7个实词后锁定正确文章，且94%-98%的多问标题首问即锁
- 数字和日期对向量影响最大，回答读取时能按序匹配子问题
本研究将2,144条来自越南法律图书馆Thu Vien Phap Luat的标题逐字输入Nemotron-3-Embed 8B/1B与Qwen3-Embedding 8B/0.6B模型，编码共65,444个前缀，匹配20,034篇文章，以及1,112个多问标题下的3,438个子问题和168个答案。结果显示：（i）黄金文章在6-7个内容词后平均进入首位，78%-85%情况下保持到底；（ii）多问标题中，首问94%-98%情况下率先锁定，第二问89%-95%不改变排名，单独编码时仅42%-58%达首位，而首问为91%-96%；（iii）数字、日期和工具标识符对嵌入的影响是内容词的2倍、诘问词的4倍，72%-78%步长朝向黄金文章，结尾疑问结构则在95%-99%情况下逆向移动；（iv）聚类发现六类典型轨迹（即时、典型、不稳定、延迟、永不锁定），不同法律领域差异显著（卡方检验p < 1e-8），房地产与诉讼类标题从不因数字锁定，环境与会计类则三分之一情况会锁定；（v）逐字读取答案后，8-16词内可召回对应文章，83%-89%按提问顺序回应子问题；（vi）单个词的移动方向在不同标题间具一致性（余弦0.25-0.33；数字达0.44-0.60），前置问题使该方向旋转约60度，问候语旋转约30度，步长随索引衰减为i^{-0.8}，双问标题整体路径接近两问线性混合（夹角12-17度），我们称之为上下文调制加法行走。
原文摘要 · Abstract (English)
A dense retriever encodes a question as one vector, but the question arrives one word at a time. We read 2,144 held-out headlines from Thu Vien Phap Luat (Vietnamese legal library) word by word with Nemotron-3-Embed 8B/1B and Qwen3-Embedding 8B/0.6B, encoding 65,444 prefixes against 20,034 articles, plus every prefix of 3,438 sub-questions from 1,112 multi-question headlines and of 168 answers. (i) The gold article becomes rank 1 after a median of 6-7 content words in every encoder, before the interrogative frame is read, and stays there to the end in 78-85% of cases. (ii) In a multi-question headline the lock is inside the first sub-question 94-98% of the time; the second leaves rank unchanged in 89-95%; encoded alone, the second reaches rank 1 in 42-58% vs 91-96% for the first, at the same lock word (95-97% identical). (iii) Numbers, dates and instrument identifiers move the embedding twice as far as content words and four times as far as interrogative words; 72-78% of steps move toward the gold article, and the closing interrogative frame moves against that direction in 95-99% of headlines. (iv) Rank/cosine clustering yields six archetypes (instant, typical, unstable, late, never-locking) that differ by legal area and form (chi-squared p < 1e-8): real-estate and litigation headlines never lock on a number; environmental and accounting headlines do so a third of the time. (v) An answer read word by word retrieves its article after 8-16 words and addresses the sub-questions in order asked in 83-89% of cases. (vi) A word's step keeps a consistent direction across headlines (cosine 0.25-0.33; 0.44-0.60 for numbers); a preceding question rotates that step by about 60 degrees and a greeting by about 30 degrees; steps shrink as i^{-0.8}; and a two-question headline is within 12-17 degrees of a linear mix of its two questions. We call this a context-modulated additive walk.
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。