arXiv cs.CL· Yuhao Chen, Shuochen Liu, Jiayao Shi, Jian Hong, Chen Cheng, Xinyun Ding, Tao Wang, Ya Li, Quan Liu, Tong Xu·· 9 天前AI 评分23
DPPM：面向个性化语言模型的双路径参数化记忆
DPPM: Dual-Path Parametric Memory for Personalized Language Models
AI 导读
针对跨会话个性化中独立上下文编译无法整合历史、循环更新会削弱早期证据的问题，研究者提出双路径参数化记忆 DPPM：Evidence 路径直接池化交互历史表示以保留早期证据，Delta 路径顺序更新关联状态以捕捉变化，融合后生成以历史为条件的 LoRA 适配器。在多个骨干模型上，DPPM 于 PersonaMem-v2 取得 54.22%、PrefEval 取得 86.79%，优于所评估的基线。
正文 · 原文
Abstract:Long-term personalization requires language models to use interaction history to track users' preferences across sessions. Parametric memory encodes this interaction history into model parameters or adapters, reducing the need to include it in the inference context. However, independent context compilation leaves cross-session integration unspecified, while recurrent updates can attenuate earlier evidence. To address these challenges, we propose Dual-Path Parametric Memory (DPPM). Its Evidence path directly pools representations of the interaction history to preserve earlier evidence, while its Delta path sequentially updates an associative state to capture changes. Fusing both outputs produces history-conditioned LoRA adapters that combine evidence accumulation with ordered revision. Across multiple backbones, DPPM outperforms the evaluated baselines, achieving 54.22% on PersonaMem-v2 and 86.79% on PrefEval. These results suggest that DPPM provides a simple and effective design choice for cross-session personalized parametric memory.
| Comments: | 12 pages, 5 figures |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.11776 [cs.CL] |
| (or arXiv:2610.11776v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.11776 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yuhao Chen [view email]
[v1]
Thu, 8 Oct 2026 11:56:07 UTC (1,811 KB)
来源：arXiv cs.CL · arxiv.org