Abstract: The high-quality supply of training data plays a decisive role in promoting the development of large-scale model technology. However, the current legal regulation for large-scale model training data in China focuses primarily on security control and risk constraints, relatively neglecting the incentive function of the supply side. This has led to a structural dilemma of triple failure in source generation, market regulation, and government intervention. To promote the continuous and stable supply of the high-quality training data, based on the symbiotic logic of "effective market" and "active government", incentive-based legal regulation can be introduced into the legislation on the supply of large-scale model training data. This can transform behavioral constraints into behavioral incentives through incentive-compatible mechanisms. Specifically, the principles of incentive-based legal regulation featured with value balance, hierarchical coordination, and appropriate proportion should be established. The scope of subjects and objects of incentive-based regulation should be broadened to form a diversified and collaborative incentive mechanism, and incentive-based regulatory tools such as rights-granting, benefit-cost, and qualification-honoring types should be improved, strengthening legislative support and application coordination. Simultaneously, under the premise of incentive-based regulation that respects market rules, appropriate mandatory regulation should be supplemented to achieve a balance between market regulation and government intervention.