🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Hypencoder: Hypernetworks for Information Retrieval Introduces a novel neural network architecture that produces small query-dependent networks for ranking, while being fast enough to search 8.8M documents in 60ms. 📝arxiv.org/abs/2502.05364 👨🏽‍💻github.com/jfkback/hypenc…

I published Vol. 91 of "Top Information Retrieval Papers of the Week" on Substack. 🔗 recsys.substack.com/p/adding-reaso…

Scaling Sparse and Dense Retrieval in Decoder-Only LLMs Hansi Zeng et al investigate how different retrieval paradigms scale with larger models, showing sparse retrieval consistently outperforms dense retrieval while demonstrating better generalization 📝arxiv.org/abs/2502.15526

Scaling Sparse and Dense Retrieval in Decoder-Only LLMs: arxiv.org/abs/2502.15526 Study scaling behaviors of retrieval models with different retrieval paradigms (sparse vs. dense) and fine-tuning objectives (contra. loss vs. KD) from 1B to 8B scale. Code: github.com/HansiZeng/scal…

Excellent video explaining Hypencoder for those who haven’t read the paper yet! Great work vishal!