Dr-DCI：通过动态工作空间扩展来实现直接语料库交互的扩展
Dr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace Expansion
摘要
在大规模语料库上的搜索过程中，通常会依赖一些基于检索器的接口来实现候选文档的筛选。这些接口在排序相关文档方面效果显著，但它们只能将查询结果以排序后的形式呈现出来，或者仅提供有限的文档信息，这限制了用户重新组织材料以及验证不同文档之间关系的能力。而直接对语料库进行操作的方法则能够克服这一限制，因为它允许用户灵活地执行搜索、过滤、比较和验证等操作。不过，随着语料库规模的增加，这种方法的性能会下降，因为需要处理的命令数量也会随之增加。我们提出了DR-DCI框架，该框架将检索过程视为一种可调用的动作，从而扩展局部工作空间。与直接处理整个语料库相比，该框架能够动态地将相关文档引入到局部工作空间中，并在其中进行搜索操作。这种设计既保持了检索的扩展性，又保留了有效处理证据所需的局部操作能力。实验表明，DR-DCI在不同规模上都能保持高效和准确。在Browsecomp-Plus测试集上，DR-DCI的准确率达到了71.2%，比单纯的DCI或改进后的版本提升了8.3个百分点，同时还减少了工具的使用次数、处理时间以及成本。通过保留工作空间的上下文重置功能，DR-DCI的准确率进一步提升到73.3%。在语料库规模从10万到1000万文档的情况下，DR-DCI仍然表现出良好的性能；而单纯的DCI则变得不稳定，BM25的性能也大幅下降。此外，DR-DCI还能适用于文件数量为2000万的场景，在Wiki-18 QA测试集上的平均得分达到了63.0，远远超过了基于检索的搜索方式以及经过训练的搜索代理模型。消融分析进一步表明，排序后的预览功能以及跨文档的搜索功能对于提升性能至关重要。
English Abstract
Agentic search over large corpora relies on retriever-mediated interfaces (e.g., BM25 or ColBERT) for scalable candidate discovery. While effective at ranking relevant documents, these interfaces expose evidence only as ranked results or bounded document views, limiting agents' ability to reorganize material and verify constraints across documents. Direct Corpus Interaction (DCI) addresses this limitation by exposing shell-executable corpus operations for flexible search, filtering, comparison, and verification. However, full-corpus terminal commands become slow and unstable as the corpus grows, degrading performance and efficiency. We introduce DR-DCI, a retriever-steered DCI framework that treats retrieval as an agent-callable action for expanding a local workspace. Rather than operating directly over the full corpus, the agent dynamically pulls relevant documents into an evolving workspace and conducts DCI operations within it. This design combines retriever-level recall with DCI-style precision: retrieval keeps exploration scalable, while DCI preserves the local operations needed for effective evidence resolution. Experiments show that DR-DCI is both effective and efficient across scales. On Browsecomp-Plus, DR-DCI reaches 71.2\% accuracy, improving over raw DCI and ablated variants by up to 8.3 points while reducing tool usage, wall time, and estimated cost. With workspace-preserving context reset, accuracy further improves to 73.3\%. In corpus-scaling experiments, DR-DCI remains effective from 100K to 10M documents, whereas raw DCI becomes unstable and BM25 performs substantially worse. DR-DCI also scales to a 20M-scale file-per-document Wiki-18 QA setting, achieving an average score of 63.0 across six benchmarks and outperforming retrieval-based and trained search-agent baselines. Ablation analysis further shows that ranked previews and inter-document DCI are key to performance.