Paper 2606.12087
FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents
- Published
- Jun 2026
- Research lab
- Independent
- Citations
- 1
- GitHub
- 27 stars
01 In brief
Summary
The paper introduces FORT-Searcher, a deep search agent trained using FORT, a framework for synthesizing shortcut-resistant training data.
The authors argue that existing methods for creating deep search tasks often fail to induce real search difficulty because agents can exploit shortcuts, such as evidence co-coverage, single-clue selectivity, exposed constants, and prior-knowledge binding.
They formalize a difficulty framework and propose FORT to control these risks during data synthesis.
FORT involves four stages: graph initialization (selecting long-tail entities and using cycle structures), graph construction (collecting diverse facts and constructing derived facts), question formulation (withholding names and fuzzing values), and adversarial refinement.
Experiments show that FORT-generated data leads to longer pre-answer search and fewer shortcut patterns compared to existing datasets.
FORT-Searcher, trained with supervised fine-tuning only, achieves the best overall performance among comparable-size open-source search agents on benchmarks like BrowseComp and BrowseComp-ZH.
02 From the paper
Abstract
Training deep search agents requires verifiable questions whose answers remain unavailable until sufficient evidence has been acquired through search. Existing synthesis methods often increase apparent difficulty by enriching graph structures, but structural complexity alone does not guarantee realized search difficulty: the intended search process can collapse through a cheaper identifying route. We formalize this gap with a shortcut-aware difficulty framework and identify four actionable shortcut risks: evidence co-coverage, single-clue selectivity, exposed constants, and prior-knowledge binding. To diagnose their realized effects, we use trajectory signatures including solving cost, answer hit time, and prior-shortcut rate. Guided by this framework, we introduce FORT, a Framework of Shortcut-Resistant Training-Data Synthesis. FORT constructs shortcut-resistant training data by controlling shortcut risks across entity selection, evidence graph construction, question formulation, and adversarial refinement. Experiments show that FORT induces longer pre-answer search and fewer shortcut patterns than existing open-source deep search datasets. Using the resulting trajectories, we train FORT-Searcher with supervised fine-tuning (SFT) only, and it achieves the best overall performance among comparable-size open-source search agents on challenging deep search benchmarks. Relevant resources will be made available at https://github.com/RUCAIBox/FORT-Searcher.