Thank you, Yuan. So starting with your first question on freight brokerage. Our freight brokerage business maintained stable operations in the first quarter with ongoing improvements to both business structure and operating model. Beginning in this year, the business has formally transitioned into a dual track model, operating its own proprietary business in parallel with aggregator model. Under the self-operated model, an extension of traditional freight brokerage business with revenue recognized on the freight brokerage business service item, FTA directly manages invoicing and settlement workflows. This model primarily serves SME shippers with genuine freight demand by providing a fully integrated end-to-end solution that combines VAT invoices issuance with freight matching. Operations have continued at their established pace with take rate or service fee remaining stable at around 10%. Under the aggregator model, this is a newly introduced track with revenue recognized under value-added services segment that invoicing and settlement are handled by qualified third-party ecosystem partners, while FTA focuses on the underlying freight matching and capacity allocation, earning a channel service fee of roughly 1% to 2% per order. This effectively repositioned the invoicing business from a GMV-driven model where the platform previously assumed full invoicing and settlement obligations to an SLI channel distribution model. From an operational standpoint, the decline of self-operated invoicing volume is the near-term outcome of our deliberate decision to reduce our self-operated exposure amid the evolving policy environment. From an asset quality perspective, the customers we retained under this model remain predominantly SOE shippers with genuine freight demand with the invoicing plus freight matching orders representing the substantial majority of the transactions. Meanwhile, the aggregator model has ramped up steadily since the first quarter launch with associated revenue beginning to flow through under the value-added services. Strategically, the transition to a dual track self-operated and aggregator model delivers 3 distinct benefits. First, it materially reduces direct exposure to regulatory policy risk. Under the aggregator model, the platform no longer bears direct invoicing and settlement obligations and fundamentally mitigating uncertainties from potential policy changes. Second, it enables an asset-lighter operating profile and sharpens our focus on core freight matching capabilities while reducing both capital deployment and operating costs. Thirdly, it strengthens shipper retention. By leveraging aggregator partners to meet shippers' invoicing compliance needs, we are better positioned to keep users engaged within our freight matching ecosystem. Financially, the invoicing business was never intended to be a core profit center. Rather, it serves as an operational infrastructure that anchors shipper loyalty and broaden the boundaries of our ecosystem. What we prioritize is the boost from the freight brokerage business to our core freight matching activity and the structural improvement it brings to our user mix. Looking ahead, we will continue to gradually transition the freight brokerage business away from the self-operated model towards the aggregator model. This shift will ensure shippers' invoicing needs are continuously served while enabling the invoicing business to operate on a lighter, more sustainable footing within the evolving regulatory environment and better supporting the long-term development of our core platform business. So that's the response to your first question. Moving on to your question on AI. In the first quarter, our AI initiatives advanced from exploratory phase to a stage of targeted capability refinement and focused testing. Centered on the core shipper transaction journey, we are progressively building an AI agent framework spanning the full transaction and fulfillment life cycle and encompassing dedicated agents for shipment posting, freight matching and other fulfillment alongside with AI-powered customer service. Our key developments in the quarter were concentrated across the following product lines. For the shipment posting agent, we continue to build on last quarter's strategy around simplified posting and automated dispatch. We steadily expanded the pilot among direct shippers, sustaining a high end-to-end success rate. Pilot results show that fulfillment rates on AI-assisted posting were materially above average. That's a strong testament to the power of AI-driven matching and improving fulfillment efficiency. Looking ahead, we plan to introduce multimodal capabilities such as screenshot-based posting to further streamline the posting experience while integrating WeCom and open APIs to meet enterprise system integration needs and improve posting efficiency. From our matching and fulfillment agents -- for our matching and fulfillment agents, core underlying capabilities went live in the first quarter. And since then, we have continue to refine their performance across intelligent query resolution, price negotiation and complex scenario handling. The matching agent focuses on dynamic negotiation strategies across varying transaction scenarios alongside growing real-time voice interaction capabilities. This fulfillment agent centers on shipment tracking, intelligent customer support and deep intervention in high-frequency exceptions such as late arrivals and cancellations and steadily establishing an automated exception handling mechanism across the platform. On the trucker side, our AI assistant continued to support high-frequency decision points such as freight finding and price negotiation and improving matching efficiency for truckers and unlocking latent capacity on the platform. Meanwhile, we continue to improve issue resolution efficiency and response speed within our AI-powered customer service system, driving structural improvements in both overall service quality and operating [ expenses ]. Looking ahead, we believe AI will continue to serve as the core technology foundation for improving operational efficiency and user experience across our platform. As we continue to refine our matching and fulfillment agents, we are also deepening the integration of our underlying models with the platform's high-frequency real-world transaction data, enabling AI to unlock greater value across matching efficiency, operating cost optimization and user experience. Thank you.