First move retrieval boosts accuracy in deep question answering agents
Question's Gambit: The First Move Matters in Agentic Deep Search
Artificial IntelligenceComputation and LanguageInformation Retrieval
Summary
Some computer programs answer hard questions by searching for information repeatedly and thinking about it. The authors noticed that the very first search these programs do can change how well they find useful info later. They designed a new way to break down the question into smaller clues to make better first searches, which helps gather better evidence and check final answers. Their tests showed this new method makes the programs more accurate at answering complex questions.
What this means in practice
- •For natural language processing engineers: Improve information retrieval modules in AI agents to boost accuracy on complex multi-step questions by carefully designing their first retrieval query.
- •For enterprise knowledge management teams: Deploy refined search strategies in automated research assistants to get more reliable and accurate answers for internal knowledge discovery tasks.
Authors
Radin Hamidi Rad, Amin Bigdeli, Negar Arabzadeh, Sajad Ebrahimi, Charles L. A. Clarke, Benjamin C. M. Fung, Ebrahim Bagheri
Abstract
Deep research agents answer complex questions through iterative loops of searching, reading, and reasoning. Recent work on reasoning-intensive benchmarks such as BrowseComp-Plus shows that well-configured lexical retrieval can surface high-quality evidence, yet agents may still fail to connect documents carrying evidence to the gold documents. We identify a deep research agent's first retrieval move as an important design decision for this setting. We introduce Question's Gambit, a first-move retrieval module that decomposes the question into a set of clues, reformulates them into complementary searches, consolidates the retrieved results, and reranks the candidate pool before the agent begins its iterative search-and-reasoning process. This produces an opening context designed to support both clue aggregation and final-answer verification. We further evaluate on MultiHop-RAG to test whether these benefits transfer beyond BrowseComp-Plus to a more conventional multi-hop question structure. Experiments on BrowseComp-Plus show that Question's Gambit improves retrieval recall and downstream agent accuracy over strong baselines, improving answer accuracy from 83.1% to 90.5% with gpt-5.5 over Pi-Serini, the strongest reported agentic baseline. Our results confirm that effective agentic deep research depends not only on the tools available inside the loop, but also on the quality of the first move. We published our implementation publicly at https://github.com/radinhamidi/Question-s-Gambit.