Researchers have introduced SWIM (Step-Wise Integrated Measure), a novel list-level evaluator designed for generative re-ranking in recommender systems. Unlike traditional methods that score lists independently, SWIM models user behavior as a session-level survival process, accounting for contextual dependencies and diminishing utility. This approach utilizes a causally-masked Transformer to efficiently estimate continuation probabilities and utilities, meeting industrial latency requirements. Experiments show SWIM significantly improves recommendation engagement compared to existing baselines. AI
IMPACT Enhances recommender system performance by better modeling user session dynamics, potentially improving engagement metrics.
RANK_REASON The item is an academic paper detailing a new method for evaluating recommender systems. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
- alphaXiv
- arXiv
- CatalyzeX
- Connected Papers
- CORE Recommender
- DagsHub
- Generator-Evaluator (G-E) framework
- Gotit.pub
- Hugging Face
- Influence Flower
- Litmaps
- ScienceCast
- scite Smart Citations
- SWIM
- Transformer++
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →