AlpsBench: An LLM Personalization Benchmark for Real-Dialogue Memorization and Preference Alignment

2026-03-09

AI Summary

Current large language models lack high-quality evaluation benchmarks grounded in real-world dialogues, hindering effective assessment of their capabilities in personalized memory and preference alignment. To address this gap, this work introduces the first structured memory evaluation benchmark built upon WildChat, a dataset of authentic long-term human–AI conversations. The benchmark encompasses four core tasks—memory extraction, updating, retrieval, and utilization—and incorporates both explicit and implicit personalization signals. Through human-validated memory annotations, a multi-task evaluation framework, and a distractor-based testing mechanism, the study systematically evaluates model performance across the entire memory lifecycle. Experimental results reveal significant limitations in state-of-the-art models, particularly in implicit feature extraction, memory update capacity, robustness to interference during retrieval, and generation of responses aligned with user sentiment.

Abstract

As Large Language Models (LLMs) evolve into lifelong AI assistants, LLM personalization has become a critical frontier. However, progress is currently bottlenecked by the absence of a gold-standard evaluation benchmark. Existing benchmarks either overlook personalized information management that is critical for personalization or rely heavily on synthetic dialogues, which exhibit an inherent distribution gap from real-world dialogue. To bridge this gap, we introduce AlpsBench, An LLM PerSonalization benchmark derived from real-world human-LLM dialogues. AlpsBench comprises 2,500 long-term interaction sequences curated from WildChat, paired with human-verified structured memories that encapsulate both explicit and implicit personalization signals. We define four pivotal tasks - personalized information extraction, updating, retrieval, and utilization - and establish protocols to evaluate the entire lifecycle of memory management. Our benchmarking of frontier LLMs and memory-centric systems reveals that: (i) models struggle to reliably extract latent user traits; (ii) memory updating faces a performance ceiling even in the strongest models; (iii) retrieval accuracy declines sharply in the presence of large distractor pools; and (iv) while explicit memory mechanisms improve recall, they do not inherently guarantee more preference-aligned or emotionally resonant responses. AlpsBench aims to provide a comprehensive framework.

Problem

• LLM personalization
• real-dialogue memorization
• preference alignment
• evaluation benchmark
• memory management

Innovation

• personalization benchmark
• real-dialogue memorization
• preference alignment
• memory management
• large language models