RecGPT-V3 技术报告

摘要

大型语言模型（LLMs）正将推荐系统从匹配历史行为中的共现模式，转向对驱动行为的意图进行推理。RecGPT-V1 通过聚焦用户理解，在淘宝上开创了这一范式；RecGPT-V2 则通过协调多智能体推理将其规模化；两者均已部署至生产环境，并在用户体验和商业成果上持续取得正向收益。然而，大规模运行 RecGPT 暴露出三个挑战：（1）无状态行为建模——每次请求都重新处理完整的用户历史，浪费计算资源且丢弃了先前的分析结果；（2）标签到物品的信息瓶颈——自然语言标签在用户理解与物品落地之间构成了有损通道；（3）低效的显式推理——冗长的思维链导致不可接受的延迟和计算开销。我们提出 RecGPT-V3，一种有状态、混合模态的推荐系统，它结合了用于开放世界知识的自然语言推理和用于具体物品落地的语义ID（SIDs）。一个记忆枢纽（Memory Hub）维护着结构化且持续演进的用户记忆，将长期行为提炼为紧凑单元，将用户建模的计算量削减 55.8%。混合模态基础模型（Hybrid-modal Foundation Model）使 LLM 能够联合推理文本标签和语义ID，开辟出一条通往物品空间的高带宽通道。潜在意图推理（Latent Intent Reasoning）将冗长的推理过程内化为紧凑的可学习潜在令牌，同时仍可解码为可读的解释，将输出令牌成本降低 200 倍。在淘宝“猜你喜欢”信息流中部署后，RecGPT-V3 在大规模在线 A/B 测试中取得了持续提升：IPV +1.28%、CTR +1.00%、TC +1.97%、GMV +3.97%，同时将端到端服务资源消耗降低了 52.4%。

English

Large language models (LLMs) are transforming recommender systems from matching co-occurrence patterns in historical behavior toward reasoning about the intent that drives it. RecGPT-V1 pioneered this paradigm on Taobao by centering user understanding, and RecGPT-V2 scaled it via coordinated multi-agent reasoning; both are deployed in production with consistent gains in user experience and commercial outcomes. However, operating RecGPT at scale reveals three challenges: (1) stateless behavior modeling, where each request reprocesses full user history, wasting computation and discarding prior analysis; (2) a tag-to-item information bottleneck, where natural-language tags form a lossy channel between user understanding and item grounding; and (3) inefficient explicit reasoning, whose lengthy chain-of-thought incurs untenable latency and compute overhead. We present RecGPT-V3, a stateful, hybrid-modal recommender that reasons over natural language for open-world knowledge and Semantic IDs (SIDs) for concrete item grounding. A Memory Hub maintains structured, continually evolving user memory that distills long-horizon behavior into condensed units, cutting user-modeling computation by 55.8%. A Hybrid-modal Foundation Model allows the LLM jointly reason over text tags and SIDs, opening a high-bandwidth channel into the item space. Latent Intent Reasoning internalizes verbose rationales into compact learnable latent tokens that remain decodable into readable explanations, lowering output token cost by 200x. Deployed in Taobao's "Guess What You Like" feed, RecGPT-V3 achieves consistent gains in large-scale online A/B tests: IPV +1.28%, CTR +1.00%, TC +1.97%, GMV +3.97%, while cutting end-to-end serving resource consumption by 52.4%.