面向未来协助需求预测的 ELBP 证据驱动小语言模型行为预测架构
Evidence-Grounded SLM Behavioral Prediction Architecture for Future Assistance Use
AI 导读
研究提出证据驱动本地行为预测架构 ELBP，用小语言模型在端侧预测用户未来的协助使用需求。在 3B 参数及以下的指令微调模型中，Qwen2.5-3B 搭配 ELBP 取得全量 F1 0.80、PR-AUC 0.84，较原始事件提示相对提升 15.9%，较 XGBoost 提升 17.6%，全量 ECE 相对降低 43.2%。
正文 · 原文
Abstract:Users often exhibit changes in interaction behavior before explicitly seeking assistance, enabling systems to anticipate future assistance use. This is especially important in high-stakes environments, where delayed assistance can have operational, financial, or safety consequences. However, conventional Large Language Models (LLMs) are often unsuitable for continuous on-device behavioral prediction because of their computational and memory demands. Behavioral patterns are also highly individual and device-specific, motivating local processing that preserves privacy and data sovereignty. Small Language Models (SLMs) offer a promising alternative, but their ability to reason reliably over low-level sequential interaction events remains underexplored. This study introduces Evidence-Grounded Local Behavioral Prediction (ELBP), a four-layer architecture combining deterministic interaction processing, dual-view behavioral representation, SLM-based prediction, and evidence verification with abstention. Three instruction-tuned models of 3B parameters or fewer were evaluated using raw, structured, dual-view, and complete ELBP configurations. The best-performing model, Qwen2.5-3B with ELBP, achieved a full-set F1-score of 0.80 and PR-AUC of 0.84, representing a 15.9\% relative improvement over raw-event prompting and 17.6\% over XGBoost, while full-set ECE decreased by 43.2\% relative to raw-event prompting. With evidence verification and abstention, selective F1 increased to 0.84 at 91\% coverage and evidence support reached 96\%. These results show that structured behavioral representation and evidence-verified selective prediction can improve the reliability of behavioral forecasting in SLMs.
| Comments: | Accepted for publication at the 2027 International Conference on Artificial Intelligence and Applied Intelligent Systems (AIAIS 2027), to be held in Hong Kong from 22 to 24 January 2027 |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.32193 [cs.AI] |
| (or arXiv:2610.32193v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.32193 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Adeyemi Adeseye Mr. [view email]
[v1]
Fri, 23 Oct 2026 05:39:31 UTC (4,186 KB)
来源：arXiv cs.AI · arxiv.org