Reproduce “BloomIntent: Automating Search Evaluation with LLM-Generated Fine-Grained User Intents” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yoonseo Choi, Eunhye Kim, Hyunwoo Kim, Donghyun Park, Honggu Lee, Jinyoung Kim, Juho Kim
A user-centric search evaluation method, BloomIntent, employs Large Language Models to generate and assess diverse, fine-grained user intents for search queries. The system's two-stage pipeline for intent generation and contextualized evaluation demonstrated evaluable, realistic intents and moderate agreement with human experts.
Key insights
- LLMs can effectively generate diverse, evaluable, and realistic fine-grained user intents, often uncovering goals beyond typical human imagination or biases.
- Automated LLM-based evaluation of SERPs against specific intents provides reasonable agreement with human judgments, particularly for clear-cut cases, serving as a cost-effective diagnostic tool.
- Clustering of intents is crucial for reducing review burden and enables search specialists to derive actionable insights at a higher, more manageable level.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.