arXiv cs.CL· Christoffer L\"offler·· 23 小时前AI 评分20
长尾多标签分类中的标签依赖:尾部标签的召回有多少靠“同伴”撑起
When the tail is earned by its companions: measuring label dependence in long-tailed multi-label classification
AI 导读
研究提出一种训练前即可计算的统计量,用于界定长尾多标签分类中报告的召回率能比“无同伴”召回高出多少。在 23 个语料库、7 种训练目标和五类标签相关机制上,所有 6 个可定义该指标的语料库中无同伴召回均不达标,最多 18.4% 的尾部真阳性在无同伴召回率下会消失。现有长尾目标均未降低这种依赖,但模型架构会改变它:尾部得分由其他标签预测算出的模型依赖显著更强。
正文 · 原文
Abstract:A rare label seldom occurs alone, and a classifier can recover it from the frequent labels it co-occurs with, its companions, without having learned the label itself. Existing diagnostics of this need a trained model: none says before training whether a corpus admits the failure, or what the standard aggregate hides. Macro-F1, like any measure built from per-label confusion counts, is unchanged by how a label's recall divides between occurrences with and without a companion. We introduce a within-label estimand of that division, a reading rule under which a fall in dependence counts only if companion-free recall rises, and a statistic of training labels that bounds, before training, how far reported recall can exceed companion-free recall. Across 23 corpora, 7 training objectives and five families of label-correlation mechanisms, companion-free recall falls short on all six corpora where the estimand is defined. Up to 18.4% of the tail true positives macro-F1 credits would vanish at the companion-free rate, within the training-label bound on every corpus. No long-tail objective reduces the dependence, but architecture moves it. Models whose tail score is computed from other labels' predictions are strongly dependent.
| Comments: | 27 pages, 3 figures, 13 tables. Submitted to Information Sciences |
| Subjects: | Machine Learning (cs.LG); Computation and Language (cs.CL); Information Retrieval (cs.IR) |
| Cite as: | arXiv:2611.22859 [cs.LG] |
| (or arXiv:2611.22859v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.22859 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Christoffer Loeffler [view email]
[v1]
Tue, 17 Nov 2026 14:38:19 UTC (127 KB)
来源:arXiv cs.CL · arxiv.org