跳到正文
HuggingFace Daily Papers(社区热门论文)
47AI 编辑部评分,满分 100

GLIE:面向视觉文档检索的生成式后期交互嵌入

2026-09-10 08:00· 1天前
AI 导读

针对视觉文档检索中后期交互嵌入存储开销大的问题,研究者提出生成式后期交互嵌入 GLIE:利用向量恰好落在单位球面、且集中在内在维度 5-6 的流形上的几何特性,每页仅用 k 个归一化质心向量作为索引,查询时只检索这 k 个向量,再由解码器将 top 候选还原为全部 N 个向量做精确重打分。

Abstract:Late-interaction retrieval is the state-of-the-art for visual document search, but it pays for its accuracy in storage. Existing compression methods retain a subset or local average of the N~1,000 vectors per page. Under aggressive storage budgets, however, these methods degrade sharply, and alternatives require retraining the encoder. Investigating this degradation across three encoders, we found two consistent properties: the vectors lie exactly on the unit sphere and concentrate near a manifold of intrinsic dimension five to six. This geometry yields two insights. First, standard k-means centroids fall inside the sphere, causing systematic underestimation of MaxSim scores. Normalizing them to the surface is a free correction worth up to +0.093 nDCG@5 over raw centroids. Second, because the page manifold has few degrees of freedom, the full set of vectors can be regenerated from only a few. To this end, we introduce Generative Late-Interaction Embeddings (GLIE): k << N vectors per page learned from the normalized centroids to serve as both a lightweight index and a basis for regenerating the page's full embedding set. At query time, search runs exclusively on these k vectors, and a decoder expands only the top candidates back to all N vectors for exact rescoring. At four vectors per page on ViDoRe v1, GLIE retains nearly 80% of the uncompressed system's nDCG@5, against 70% for the best prior post-hoc method. These results use a 415K-parameter network fitted in under three GPU-minutes on just a thousand training pages. At a matched training budget, fine-tuning the encoder does not reach even the training-free stage of GLIE, and the full system beats it at every budget. These patterns hold across a second encoder and ViDoRe v2. By reconstructing evidence on demand rather than sampling it, GLIE opens a new axis for storage-efficient retrieval, with the decoder as its main design surface.
Subjects: Information Retrieval (cs.IR)
Cite as: arXiv:2609.11808 [cs.IR]
  (or arXiv:2609.11808v1 [cs.IR] for this version)
  https://doi.org/10.48550/arXiv.2609.11808
arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Mohamed Eltahir [

Thu, 10 Sep 2026 16:58:39 UTC (2,161 KB)

Access Paper:

Additional Features

Current browse context:

References & Citations

Bookmark

Bibliographic and Citation Tools

Code, Data and Media Associated with this Article

来源:HuggingFace Daily Papers(社区热门论文)· arxiv.org