Succeeding at Scale: Enterprise Retrieval Benchmark Construction and Index-Preserving Query Adaptation for Multi-Tenant Search
arXiv:2601.04646v4 Announce Type: replace-cross Abstract: Large-scale multi-tenant retrieval systems generate extensive query logs but lack curated relevance labels for effective domain adaptation, resulting in substantial underutilized "dark data." This challenge is compounded by the high cost of model updates, as jointly fine-tuning query and document encoders requires full corpus re-indexing, which is impractical in multi-tenant settings with thousands of isolated indices. We introduce DevRev-Search, a passage retrieval benchmark for technical customer support built via a fully automated pi
The proliferation of context-aware AI systems and the increasing complexity of enterprise data environments make efficient, adaptable retrieval crucial for realizing AI's full potential.
This development addresses a key bottleneck in deploying scalable, domain-adapted AI for enterprise search, enabling more efficient utilization of vast, unstructured data within multi-tenant systems.
The proposed benchmark and adaptation methods simplify the process of fine-tuning and updating retrieval models without costly re-indexing, accelerating AI adoption in complex enterprise settings.
- · Enterprise AI providers
- · Customer support automation
- · Large language model applications
- · Data-intensive businesses
- · Legacy search solutions
- · High-cost manual data labeling services
- · Companies with undifferentiated 'dark data'
Enterprise search and knowledge management systems become significantly more performant and adaptable, reducing operational costs.
Improved information access within enterprises leads to higher productivity, better decision-making, and enhanced customer experiences.
The development of highly specialized and efficient retrieval systems accelerates the broader adoption of AI agents by providing robust, contextually relevant data access.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI