A new framework called X-split Argus has been developed for ad ranking, aiming to improve click-through rate (CTR) prediction efficiency. This system uses a compressor to encode past interactions into reusable tokens, which are then utilized by an online ranker. The approach showed a 0.81% reduction in normalized entropy compared to a baseline and resulted in a 4.65% increase in a key business metric and a 2.96% increase in clicks during an A/B test, despite a 15% rise in online inference cost. AI
IMPACT This framework could lead to more efficient and effective ad ranking systems by optimizing the use of user history data.
RANK_REASON This is a research paper detailing a new framework for CTR prediction. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
- Alexander Ploshkin
- alphaXiv
- arXiv
- CatalyzeX
- click-through rate
- DagsHub
- Gotit.pub
- Hugging Face
- ScienceCast
- X-split Argus
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →