Memory-Based vs. Context-Only Conditioning Produces Distinct Behavioral Patterns in Stateful Personalization

Junsoo Park; Youssef Medhat; Htet Phyo Wai; Ploy Thajchayapong; Ashok K. Goel

cs.IRApr 14, 2026

Memory-Based vs. Context-Only Conditioning Produces Distinct Behavioral Patterns in Stateful Personalization

Authors: Junsoo Park, Youssef Medhat, Htet Phyo Wai, Ploy Thajchayapong, Ashok K. Goel

Organizations: Georgia Institute of Technology, Atlanta, GA, USA

Abstract

We study how conditioning context shapes personalization behavior in a teacher-facing educational recommender system. We compare contextual conditioning based on the current student question with memory-based conditioning using persistent learner information. Using deviation correlation and paired statistical tests, we find that contextual recommendations exhibit stronger question-level responsiveness, while memory-based recommendations exhibit history-dependent behaviors, including learner-specific differentiation under identical input. Teacher-facing evaluation signals suggest these recommendations are interpretable and actionable. These results indicate that embedding-based similarity metrics capture responsiveness to the current question but do not characterize personalization grounded in learner history, motivating behavior-level diagnostics for studying conditioning effects.

Figures & tables

Figure 1: Memory-based recommendation pipeline (learner-conditioned). The learner state functions as a persistent memory across interactions, retrieved at each step to guide pedagogical tactic selection prior to generation. The same generation mechanism is used across conditions; memory-based conditioning introduces memory retrieval and tactic selection.

Condition	Abridged prompt inputs
Contextual (question-conditioned)	Question: {student_question} $\rightarrow$ Generate teacher-facing recommendation.
Memory-based (learner-conditioned)	Learner state: {top_need, persona}; Tactic: {feed_up/feed_back/feed_forward}; Question: {student_question} $\rightarrow$ Generate recommendation consistent with tactic.

Table 1: Abridged prompt skeletons illustrating the conditioning delta (same student question).

	Learner A	Learner B
Dominant need	Engagement	Performance
Selected tactic	Feed Up	Feed Back
Instructional focus	Goal clarity and relevance	Mastery status and accuracy feedback
Contextual output	Have the student justify their True/False choice using a claim–evidence–reasoning structure and rewrite the statement to make it testable.
Memory-based output	Clarify the purpose of the task and why evaluating claims about cognitive tasks matters in the course context; ask the student to restate the goal in their own words before proceeding.	Provide accuracy-based feedback relative to course mastery thresholds, identify the specific misconception in the response, and highlight concepts already mastered.

Table 2: Personalized recommendations under identical input. Memory-based outputs diverge across learners due to differences in persistent learner state, illustrating memory-based personalization beyond contextual conditioning.

Recommender Systems

Recommender systems predict and rank items for people using interaction histories, content, context, and feedback.

See all 508 papers

Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents
May 25, 2026Yeonjun In, Wonjoong Kim, Sangwu Park +2Memorization in Language Models Persistent Memory for Language Models
KAIST
PersonaFactor: Preference Factors with Dynamic Selection for LLM Personalization
Oct 12, 2026Hang Su, Zequn Liu, Jianghan Chao +3LLM Personalization User Preference Modeling
East China Normal University · Beijing Zhongguancun Academy · Renmin University of China +1
Learning to Reason for Multi-Step Retrieval of Personal Context in Personalized Question Answering
Feb 22, 2026Maryam Amirizaniani, Alireza Salemi, Hamed ZamaniQuestion Answering Retrieval-Augmented Generation
University of Washington Seattle, WA, USA · University of Massachusetts Amherst Amherst, MA, USA

May 25, 2026J/K move · Enter open · S save

Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents

Yeonjun In, Wonjoong Kim, Sangwu Park, Kanghoon Yoon+1

KAIST

Memorization in Language Models Persistent Memory for Language Models Memory Agent Benchmarks Long-Horizon LLM Agents Agent Memory Management

Open paper AskPDF arXiv

Existing large language model (LLM) based memory systems apply universal, static policies that overlook a fundamental reality: the contexts that are worth storing in memory are different across users. This misalignment wastes limited memory budget on transient interactions while failing to preserve critical context for long horizon tasks. To address this gap, we investigate an underexplored question: can LLM based memory systems learn personalized memory policies? We introduce PerMemBench, the first benchmark for evaluating personalized memory systems, featuring multi year, multi domain interaction histories across diverse user personas. We further present the first empirical study of memory personalization, proposing session level storage gating, a lightweight framework that selectively bypasses memory operations for transient sessions. Our study confirms that personalization yields substantial retention gains under perfect gating, yet reveals that accurate gating remains an open and critical challenge.

Abstract

Recommender Systems

Explore similar work

Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents

PersonaFactor: Preference Factors with Dynamic Selection for LLM Personalization

Learning to Reason for Multi-Step Retrieval of Personal Context in Personalized Question Answering