DPPM: Dual-Path Parametric Memory for Personalized Language Models

2026-10-08

AI Summary

This study addresses the issue of early evidence decay in cross-session personalization, which arises from context compilation independence and cyclic updating. To mitigate this problem, we propose a dual-path parametric memory mechanism that generates LoRA adapters by fusing two complementary pathways: evidence accumulation and incremental revision. This approach enables correlated state modeling, effectively balancing the retention of long-term historical evidence with the orderly updating of states across sessions. Experimental evaluations demonstrate that the proposed model achieves performance scores of 54.22% and 86.79% on the PersonaMem-v2 and PrefEval benchmarks, respectively, significantly outperforming existing baseline methods.

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.

Problem

• Personalized Language Models
• Parametric Memory
• Cross-session Integration
• Long-term Personalization

Innovation

• Dual-Path Parametric Memory
• Personalized Language Models
• LoRA Adapters
• Cross-session Personalization
• Parametric Memory