Seed2.0 模型卡：邁向應對現實世界複雜性的智能前沿

摘要

我們推出了Seed2.0模型系列，這是在解決複雜真實世界任務上邁出的重要一步。我們的做法始於辨識用戶的真實需求，並透過選取與抽象化基於這些需求、貼近現實複雜場景的基準，來建構一套可靠且具前瞻性的評估體系。在此評估體系的引導下，Seed2.0針對長尾知識與複雜指令遵循這兩項長期挑戰，大幅提升了模型在複雜、長程任務上的可靠性。除此之外，Seed2.0還提供了世界領先的推理智能、視覺理解與搜尋能力，以滿足廣大用戶最普遍的需求。透過本模型卡中記錄的大量真實世界應用案例，我們證明Seed2.0已開始展現處理初步複雜真實世界任務的能力，為數億用戶創造更大的價值。

English

We present Seed2.0, a model series that takes a meaningful step toward solving complex, real-world tasks. Our approach begins with identifying users' genuine needs and constructing a reliable, forward-looking evaluation system by selecting and abstracting benchmarks grounded in these needs and in realistic, complex scenarios. Guided by this evaluation system, Seed2.0 targets two persistent challenges, long-tail knowledge and complex instruction following, substantially improving the model's reliability on intricate, long-horizon tasks. Beyond these, Seed2.0 delivers world-leading reasoning intelligence, visual understanding, and search capabilities that address the most common needs of a broad user base. Through extensive real-world use cases documented in this model card, we demonstrate that Seed2.0 begins to exhibit the ability to handle initial complex real-world tasks, delivering greater value to hundreds of millions of users.