Cross-Domain Hybrid OPD for Generalizable Search Agents

Recent advances in Reinforcement Learning (RL) have substantially improved the capabilities of autonomous search agents, enabling sophisticated planning, and iterative retrieval over dynamic information sources. However, optimizing language models for specialized search behaviors often incurs an alignment tax, where gains in search performance come at the expense of general-purpose capabilities, limiting their effectiveness as universal assistants. In this technical report, we present the training framework behind the Yuanbao search agent, designed to achieve search specialization without sacrificing general intelligence. Built upon the Hunyuan3 architecture, our framework combines agentic reinforcement learning for autonomous search with a cross-domain expert On-Policy Distillation (OPD) pipeline. Experts specializing in complementary general-purpose domains are distilled into the search-specialized student, restoring and further enhancing its broad capabilities. Rather than treating specialization and general capability as competing objectives, our hybrid training strategy jointly optimizes both, effectively mitigating the alignment tax. Extensive experiments demonstrate that the resulting model achieves competitive search performance while consistently improving its general-purpose capabilities, providing a favorable balance between specialized execution and broad generalization in real-world search scenarios.

A genetic algorithm based superdirective beamforming method under excitation power range constraints

The array gain of a superdirective antenna array can be proportional to the square of the number of antennas. However, the realization of the so-called superdirectivity entails accurate calculation and application of the excitations. Moreover, the excitations require a large dynamic power range, especially when the antenna spacing is smaller. In this paper, we derive the closed-form solution for the beamforming vector to achieve superdirectivity. We show that the solution only relies on the data of the array electric field, which is available in measurements or simulations. In order to alleviate the high requirement of the power range, we propose a genetic algorithm based approach with a certain excitation range constraint. Full-wave electromagnetic simulations show that compared with the traditional beamforming method, our proposed method achieves greater directivity and narrower beamwidth with the given range constraints.