跳到正文
arXiv cs.AI· Ilia Karpov, Kirill Koloskov, Valeriy Kuleshov, Dmitry Ilvovsky·· 10 小时前AI 评分18

如何解读动态链接预测中的时间戳置换实验

Interpreting Timestamp Permutations in Dynamic Link Prediction

AI 导读

研究提出交叉诊断方法,将时间戳置换带来的性能差距拆分为训练条件与评估历史两种效应,并测量输入暴露程度。在 MOOC 数据集上,TCL 用二元交叉熵训练时,原始时序训练在两种评估历史下均带来约 8 个 pairwise 准确率点提升,而固定权重下改变历史仅 0.03 点,因 68.37% 查询输入未变。DyGFormer 的训练与历史对比分别为 22.34 和 0.38 点。

正文 · 原文

View PDF HTML (experimental)

Abstract:Timestamp permutations are used to test whether dynamic link predictors exploit temporal information, but a performance gap can reflect changes in training, evaluation history, or both. We separate these effects with a crossed diagnostic and measure input exposure: whether altered histories actually change the predictor's inputs. On MOOC, TCL trained with binary cross-entropy gains approximately eight pairwise-accuracy points from original-chronology training under either evaluation history, while changing history at fixed weights gives only 0.03 points. The small history contrast reflects unchanged inputs for 68.37% of queries, stable rankings for most others, and cancellation of gains and losses. Under original evaluation history, suppressing explicit time encoding reduces the training contrast from 8.03 to 3.64 points, a paired reduction of 4.39 points (95% interval [3.68, 5.09]) across six seeds. DyGFormer shows corresponding training and history contrasts of 22.34 and 0.38 points across two seeds, despite greater exposure. Partial history permutations further separate nominal intervention strength, input changes, and aggregate accuracy. These comparisons turn a single permutation gap into interpretable statements about training conditions and model responses, while leaving the internal mechanisms of temporal learning open.
Comments: 5 pages total (ACM sigconf format). Companion code and trained models: this https URL
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Logic in Computer Science (cs.LO); Social and Information Networks (cs.SI)
Cite as: arXiv:2611.23753 [cs.LG]
  (or arXiv:2611.23753v1 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2611.23753

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Ilia Karpov [view email]
[v1] Wed, 18 Nov 2026 03:08:09 UTC (73 KB)

来源:arXiv cs.AI · arxiv.org