Adding Reasoning Capabilities to Dense Retrievers, Making Scaling Laws Work in Industrial Recommender Systems, and More!

Vol.91 for Feb 10 - Feb 16, 2025

Stay Ahead of the Curve with the Latest Advancements and Discoveries in Information Retrieval.

This week’s newsletter highlights the following research:

1. Human-Guided Progressive Training for E-Commerce Ad Retrieval, from Walmart

2. A Mechanistic Interpretation of Cross-Encoder Relevance Scoring, from Lu et al.

3. A Graph-Enhanced RAG for Precise Medical Diagnosis and Treatment Recommendation, from NTU

4. A Decomposed Approach to Scaling Industrial Recommendation Models, from Alibaba

5. Experimental Lessons from Graph-Based Approximate Nearest Neighbor Search, from Azizi et al.

6. Learning Query-Dependent Neural Networks for Document Retrieval, from UMass Amherst

7. Reasoning-Augmented Dense Retrieval through Large Language Models, from Yan et al.

8. Coarse-to-Fine Tree Learning for Scalable Document Retrieval, from ServiceNow

9. mmE5: Efficient Multimodal Embeddings through Quality-Focused Data Synthesis, from Microsoft

10. A Framework for Designing Multi-Objective Ranking Models, from Amazon