Subjective Behaviors and Preferences in LLM: Language of Browsing

2025-08-21

AI Summary

This paper investigates the limitations of large language models (LLMs) in modeling users’ subjective browsing behavior, addressing three core questions: (1) whether smaller models better capture syntactically unconstrained “browsing language”; (2) whether a single-parameter model can adequately represent user heterogeneity; and (3) whether high average performance implies low individual-level variance and strong user-level alignment. To this end, we propose HeTLM—a clustering-aware heterogeneous training framework—featuring page-level tokenization, a lightweight model architecture, and user-clustering-driven heterogeneous parameter optimization. Experiments demonstrate that HeTLM significantly outperforms mainstream LLMs on browsing sequence modeling: it improves aggregate performance while substantially reducing cross-user prediction variance, thereby achieving stronger individual behavioral consistency and preference alignment. These results reveal the critical advantage of the “small-and-specialized” paradigm for user behavior modeling.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

A Large Language Model (LLM) offers versatility across domains and tasks, purportedly benefiting users with a wide variety of behaviors and preferences. We question this perception about an LLM when users have inherently subjective behaviors and preferences, as seen in their ubiquitous and idiosyncratic browsing of websites or apps. The sequential behavior logs of pages, thus generated, form something akin to each user's self-constructed "language", albeit without the structure and grammar imbued in natural languages. We ask: (i) Can a small LM represent the "language of browsing" better than a large LM? (ii) Can an LM with a single set of parameters (or, single LM) adequately capture myriad users' heterogeneous, subjective behaviors and preferences? (iii) Can a single LM with high average performance, yield low variance in performance to make alignment good at user level? We introduce clusterwise LM training, HeTLM (Heterogeneity aware Training of Language Model), appropriate for subjective behaviors. We find that (i) a small LM trained using a page-level tokenizer outperforms large pretrained or finetuned LMs; (ii) HeTLM with heterogeneous cluster specific set of parameters outperforms a single LM of the same family, controlling for the number of parameters; and (iii) a higher mean and a lower variance in generation ensues, implying improved alignment.

Problem

• Can small LMs better represent browsing behavior language?
• Can single LMs capture diverse user preferences adequately?
• Can high-performance LMs achieve low variance for user alignment?

Innovation

• Small LM with page-level tokenizer outperforms large LMs
• HeTLM uses cluster-specific parameters for heterogeneity
• Achieves higher mean and lower variance in generation