arXiv cs.CL· Zhe Xu, Tianhao Liao, Zhenqing Ling, Ying Shen, Yaliang Li, Daoyuan Chen·· 12 小时前AI 评分31
长上下文时代重思检索：免训练的语义树实现自适应上下文构建
Rethinking Retrieval in the Long-Context Era: Training-Free Semantic Trees for Adaptive Context Construction
AI 导读
研究者提出免训练框架 \ours，用语义抽象指导证据选择、保留原文用于回答，从完整文档一次性生成可复用的语义层级并关联节点描述与源文本片段。在四个前沿回答模型上，该方法于三项 LongBench 任务中平均 QA-F1 比全上下文提示提升 1.72 分，同时将回答上下文长度缩减 58.55%。代码已开源。
正文
Abstract:Long-context language models can increasingly process entire documents, making full-context prompting a practical alternative to passage retrieval. However, greater capacity determines how much information a model can receive, but not which information best supports a query. This distinction motivates rethinking retrieval: when the full document already fits, can selecting and organizing its evidence improve answering? The challenge is to retain supporting facts and relations while keeping the context focused on the question. To reconcile these goals, we introduce \ours, a training-free framework that uses semantic abstraction to guide selection while retaining original text for answering. \ours constructs a reusable semantic hierarchy from the full document in a single generation, linking node descriptions to source spans. Given a query, structural--semantic scoring and fine-grained lexical attribution nominate shared source nodes; their union with ancestor context preserves structural relations and complete local text. Across four frontier answer models, \ours improves average QA-F1 by 1.72 points over full-context prompting on three LongBench tasks, surpassing the evaluated baselines on the four-model average while reducing answer-context length by 58.55\%. Our code is available at the $\href{this https URL}{repository}$.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.34146 [cs.CL] |
| (or arXiv:2610.34146v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.34146 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhe Xu [view email]
[v1]
Sun, 25 Oct 2026 08:28:39 UTC (759 KB)
来源：arXiv cs.CL · arxiv.org