←Back to feed
🧠 AI⚪ Neutral
Succeeding at Scale: Automated Dataset Construction and Query-Side Adaptation for Multi-Tenant Search
arXiv – CS AI|Prateek Jain, Shabari S Nair, Ritesh Goru, Prakhar Agarwal, Ajay Yadav, Yoga Sri Varshan Varadharajan, Constantine Caramanis|
🤖AI Summary
Researchers introduce DevRev-Search, an automated system for building multi-tenant search systems that addresses the challenge of underutilized query data in enterprise environments. The system uses LLM-based relevance labeling and query-only fine-tuning to improve search performance while avoiding costly full corpus re-indexing.
Key Takeaways
- →Multi-tenant search systems generate extensive query logs but lack proper relevance labels, creating substantial "dark data".
- →DevRev-Search uses automated pipeline with LLM-as-a-Judge for consistency filtering and relevance labeling.
- →Index-Preserving Adaptation strategy fine-tunes only query encoders while keeping document indices fixed.
- →Parameter-Efficient Fine-Tuning (PEFT) delivers strong quality-efficiency trade-offs for enterprise search.
- →The approach enables scalable search adaptation without the impractical cost of full corpus re-indexing.
#search-systems#llm#enterprise-ai#multi-tenant#automated-labeling#query-optimization#parameter-efficient-fine-tuning
Read Original →via arXiv – CS AI
Act on this with AI
Stay ahead of the market.
Connect your wallet to an AI agent. It reads balances, proposes swaps and bridges across 15 chains — you keep full control of your keys.
Related Articles