ORBIT: 20K Reasoning-Intensive Queries for Training Search Agents Without Paid API Services
arXiv·medium signal
Introduces ORBIT, a training dataset with 20K reasoning-intensive queries with short verifiable answers, generated using a frugal framework that requires no paid API services. Addresses the bottleneck of constructing training data for deep research tasks involving multi-step retrieval and reasoning. Practical for teams building search agents who want to avoid expensive human annotation.