Researchers have developed a new pipeline for relevance evaluation in search systems using vision-language models (VLMs). This automated approach, deployed within Pinterest Search, aims to overcome the limitations of human annotation by providing faster and more scalable relevance measurements. The VLM-based system has been validated against human judgments, demonstrating its reliability and efficiency for online A/B experiments. This advancement allows for broader query sets, optimized sampling, and a more comprehensive assessment of search experiences, ultimately leading to improved relevance metrics and reduced minimum detectable effects. AI
IMPACT Enhances search relevance and efficiency by automating evaluation, potentially improving user experience and reducing development costs.
RANK_REASON The cluster contains an academic paper detailing a new methodology for relevance measurement using VLMs.
Read on arXiv cs.IR (Information Retrieval) →
- alphaXiv
- arXiv
- CatalyzeX
- Connected Papers
- CORE Recommender
- DagsHub
- Gotit.pub
- Hugging Face
- Influence Flower
- Litmaps
- Pinterest Search
- ScienceCast
- scite Smart Citations
- vision-language model
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
