arXiv cs.AI· Yang Xiao, Yusong Sun, Haoyi Wu, Wenyang Hui, Weiqiang Hang, Luqing Wang, Zejian Wang, Zirui Tu, Jingsheng Mao, Shaobo Hu, Yang Luo, Jiale Guo, Jin Zhang, Yudan Liu, Zhaokai Luo, Ren Huan, Yao Hu, Chengyue Jiang, Wenjie Li, Chuan Mu·· 3 天前AI 评分41
Vela：以用户行为为监督信号激发 LLM 智能
Vela: User Behavior as Supervision Sparks LLM Intelligence
AI 导读
研究团队提出 Vela 框架，将真实用户行为数据转化为可训练任务与轨迹，从 10 万名用户的匿名行为记录中生成 10 万个任务、覆盖 91 个真实任务模板，经难度过滤后保留约 7.5 万个任务。基于这些轨迹微调 Qwen3.6-35B-A3B，用户相关任务最高提升 11.5 分，通用基准平均提升 4 分；微调后再做强化学习，比从基座模型直接强化学习最高高出 7.7 分。
正文 · 原文
Authors:Yang Xiao, Yusong Sun, Haoyi Wu, Wenyang Hui, Weiqiang Hang, Luqing Wang, Zejian Wang, Zirui Tu, Jingsheng Mao, Shaobo Hu, Yang Luo, Jiale Guo, Jin Zhang, Yudan Liu, Zhaokai Luo, Ren Huan, Yao Hu, Chengyue Jiang, Wenjie Li, Chuan Mu
Abstract:Large language models are increasingly deployed in applications that act on users' long-term behavior histories, from personal assistants to recommendation and advertising systems. These applications demand the user dimension of LLM intelligence: inferring a person's preferences, intent, and likely future actions from weeks to years of behavior data. Yet current models are trained on static text or synthetic data, neither of which captures how user behavior evolves over time. Real-world behavior records, already available at scale on content platforms and in personal-assistant agents, offer a natural supervisory signal. We propose user behavior as supervision and introduce Vela, a framework that converts real user behavior data into trainable tasks and trajectories. From the anonymized behavior records of 100,000 users, Vela generates 100,000 tasks across 91 real-world task templates; difficulty-based filtering and trajectory acceptance criteria retain approximately 75,000 tasks with selected training trajectories. Fine-tuning Qwen3.6-35B-A3B on these trajectories improves performance over the base model by up to 11.5 points on user-related tasks and by an average of 4 points on general benchmarks. Training on more user behavior data further improves user-related performance, and reinforcement learning after fine-tuning outperforms reinforcement learning from the base model by up to 7.7 points. These gains suggest that real-world user behavior is a scalable source of supervision for both user-centered capabilities and the broader intelligence of LLMs.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.14709 [cs.CL] |
| (or arXiv:2610.14709v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.14709 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yang Xiao [view email]
[v1]
Sat, 10 Oct 2026 08:28:00 UTC (721 KB)
来源：arXiv cs.AI · arxiv.org