Retrieving Interaction Spaces for Search Agents, Breaking the Quadratic Encoder Bottleneck in Generative Retrieval, and More!
Vol.160 for Jun 08 - Jun 14, 2026
Stay Ahead of the Curve with the Latest Advancements and Discoveries in Information Retrieval.
This week’s newsletter highlights the following research:
A Bounded Middle Ground Between RAG and Direct Corpus Interaction, from Zhuang et al.
Scaling Short-Form Video Recommendation with Semantic IDs and Compressed Transformers, from Google
Training-Free LLM Embeddings via Spectral Filtering, from Wu et al.
Linear-Time Bidirectional Attention for Long-History Generative Retrieval, from Yandex
Teaching LLMs When and What to Hand Off in Deep Research, from Ning et al.
Serving LLM-Generated Interest Personas in Production Recommendation Systems, from Google
Item-Level Scoring for Generative Retrieval in Industrial Recommendation, from Yandex
Efficient Multimodal Reranking via Visual Cache Reuse and Interaction Sparsity, from Fan et al.
Grounding Retrieval in Reasoning Utility for Reinforcement Fine-Tuning, from Meta
Training Lexical Query Rewriters with Stepwise Retrieval Feedback, from Satouf et al.
