arXiv：cs.LG· Sicong Wang, Ruiting Dong, Yue Liu, Bowen Zheng, Jun Meng, Jie Li, Shuaijun Guo, Yu Gu, Fanyi Di, Xin Li·· 3 小时前
Amap 提出 GPlan：用隐式推理蒸馏实现生成式时空意图序列推荐
Generative Spatiotemporal Intent Sequence Recommendation via Implicit Reasoning in Amap
AI 导读
高德（Amap）提出生成式框架 GPlan，通过 Progressive Implicit CoT Distillation 把 LLM 显式推理压缩进保留的 latent token，使小模型在严格延迟约束下继承复杂规划逻辑，并以 Spatiotemporal Counterfactual DPO 对齐反事实上下文-规划对，减少时空上下文错配的规划。
正文
Abstract:Real-world user behavior rarely consists of isolated actions; instead, it often forms intent flows governed by spatiotemporal dependencies. To provide integrated service recommendations, we focus on the task of Generative Spatiotemporal Intent Sequence Recommendation (GSISR), which aims to generate intent sequences that are logically coherent and physically executable within complex spatiotemporal contexts. While LLMs offer strong reasoning potential for GSISR, direct industrial deployment is limited by high inference latency and context-mismatched or physically infeasible plans. To address these challenges, we propose a generative framework, GPlan, that internalizes LLM reasoning into lightweight models through two components. First, to enable reasoning under strict latency constraints, we introduce Progressive Implicit CoT Distillation, which compresses explicit reasoning processes into reserved latent tokens, allowing small models to inherit complex planning logic without generating long reasoning text. Second, to address the disconnect between general knowledge and real-world constraints, we design Spatiotemporal Counterfactual DPO. By aligning the model with counterfactual context-plan pairs, we improve sensitivity to spatiotemporal context and reduce context-mismatched plans. Offline experiments and online A/B testing demonstrate that our approach improves sequence coherence and context responsiveness. Our implementation and the anonymized GSISR dataset are available at this https URL.
| Comments: | 9 pages, 1 figure |
| Subjects: | Information Retrieval (cs.IR); Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.28888 [cs.IR] |
| (or arXiv:2605.28888v2 [cs.IR] for this version) | |
| https://doi.org/10.48550/arXiv.2605.28888 arXiv-issued DOI via DataCite |
|
| Journal reference: | Proceedings of the 20th ACM Conference on Recommender Systems (RecSys '26), pp. 1119-1127, 2026 |
| Related DOI: | https://doi.org/10.1145/3773078.3831861
DOI(s) linking to related resources |
Submission history
From: Sicong Wang [view email]
[v1]
Wed, 27 May 2026 07:27:32 UTC (169 KB)
[v2]
Thu, 8 Oct 2026 06:19:54 UTC (155 KB)
来源：arXiv：cs.LG · arxiv.org