Researchers propose Multi-Item Scoring for faster whole-page ranking
Researchers introduced Multi-Item Scoring (MIS), a new method for ranking web pages that balances efficiency and quality. Traditional non-autoregressive (NAR) approaches score items independently, missing sequential dependencies, while autoregressive (AR) methods are too slow for large-scale use. MIS combines global candidate interactions with lightweight prefix-conditioned scoring, reusing…
Key points
- MIS framework uses prefix-aware scoring to mimic autoregressive quality without full-model inference overhead
- Tests show **4x faster CPU decoding** on KuaiRand-1K and RL4RS datasets compared to autoregressive baselines
- On a social-video platform, MIS boosted watch time by **0.058%** and cut CPU use by **9.73%**
Tests on KuaiRand-1K and RL4RS show MIS matches AR quality while speeding up CPU decoding 4.07x and 3.54x, respectively. On a social-video platform, MIS improved engagement metrics by 0.104% in watch time and 0.058% in realization, while cutting CPU use by 9.73%. The paper, posted on arXiv, highlights MIS as a scalable solution for industrial whole-page ranking.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Anthropic uses Claude to map bacteriophage vault protein at 3.2 Å resolution · 3 src
- RAGCompass routes retrievers by reranking passages, improves answer F1 by 28% · 1 src
- MIT researchers develop uq-daam to correct robot memory · 1 src
- Qwen improves LLM-based OCR correction with veto gate · 1 src
- Study: LLMs prefer to answer over clarifying ambiguous references · 1 src
Comments
via GitHub Discussions