Two new research papers explore advancements in agentic search, focusing on how AI agents interact with information over extended sessions. The first paper analyzes over 14 million real search requests to understand user intents and query reformulation patterns, revealing that most sessions are short and query terms often trace back to retrieved evidence. The second paper introduces a framework for long-horizon multimodal search, addressing challenges of context management and token costs by using file-based visual representations and on-demand loading, achieving state-of-the-art results on complex benchmarks. AI
Summary written by None from 2 sources. How we write summaries →
IMPACT These papers offer insights into improving AI agent efficiency and capability in complex information-seeking tasks, potentially leading to more effective search tools.
RANK_REASON Two academic papers published on arXiv detailing new methods and analyses for agentic search systems.