PraMem: Practice-derived Experiential Memory for Long-horizon Behavior Prediction

2026-07-02

AI Summary

This work addresses the challenge that large language models struggle to effectively abstract latent behavioral patterns over long-horizon behavior prediction and are susceptible to cognitive biases, a limitation not fundamentally resolved by existing memory compression approaches. The authors propose reframing lengthy historical behavior sequences from a computational burden into a strategic resource by pre-training on practice-derived experiences to construct an experience memory, which is then integrated as auxiliary input to enhance predictive performance. Departing from conventional context compression paradigms, this framework introduces, for the first time, a practice-based experience memory mechanism. Experimental results demonstrate significant improvements over state-of-the-art methods across multiple tasks, while in-depth analysis elucidates the operational dynamics and evolutionary characteristics of the proposed experience memory.

Abstract

Long-horizon behavior prediction aims to infer a user's next action based on a lengthy historical sequence, playing a crucial role in artificial intelligence field. The rise of large language models (LLMs) offers a promising direction for sequential behavior prediction, yet LLMs struggle with latent behavioral pattern induction and model-intrinsic cognitive biases when tackling long-horizon behavior prediction. Prior memory management methods follow a context-compression paradigm that attempts to address this task by alleviating the historical sequence burden, yet fail to resolve the core challenges. In this paper, we advocate a paradigm shift that reframes the lengthy historical sequence from a burden into a valuable resource to be exploited, and accordingly propose PraMem, which conducts beforehand practice over the lengthy historical sequence to build an experiential memory, thereby serving as the assisted input for accurate long-horizon behavior prediction. Extensive experiments across diverse tasks demonstrate that PraMem achieves superior performance than prior methods, and more in-depth analyses provide valuable insights into the mechanism and evolution of the experiential memory. Code: https://github.com/icip-cas/PraMem.

Problem

• long-horizon behavior prediction
• large language models
• behavioral pattern induction
• cognitive bias
• historical sequence

Innovation

• experiential memory
• long-horizon behavior prediction
• practice-derived memory
• paradigm shift
• large language models