Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Breaking User-Centric Agency: A Tri-Party Framework for Agent-Based Recommendation

Yaxin Gong, Chongming Gao, Chenxiao Fan, Wenjie Wang, Fuli Feng, Xiangnan He

TL;DR

This work proposes the first Tri-party LLM-agent Recommendation framework (TriRec) that explicitly coordinates user utility, item exposure, and platform-level fairness and finds that item self-promotion can simultaneously enhance fairness and effectiveness, challenging the conventional trade-off assumption between relevance and fairness.

Abstract

Recent advances in large language models (LLMs) have stimulated growing interest in agent-based recommender systems, enabling language-driven interaction and reasoning for more expressive preference modeling. However, most existing agentic approaches remain predominantly user-centric, treating items as passive entities and neglecting the interests of other critical stakeholders. This limitation exacerbates exposure concentration and long-tail under-representation, threatening long-term system sustainability. In this work, we identify this fundamental limitation and propose the first Tri-party LLM-agent Recommendation framework (TriRec) that explicitly coordinates user utility, item exposure, and platform-level fairness. The framework employs a two-stage architecture: Stage~1 empowers item agents with personalized self-promotion to improve matching quality and alleviate cold-start barriers, while Stage~2 uses a platform agent for sequential multi-objective re-ranking, balancing user relevance, item utility, and exposure fairness. Experiments on multiple benchmarks show consistent gains in accuracy, fairness, and item-level utility. Moreover, we find that item self-promotion can simultaneously enhance fairness and effectiveness, challenging the conventional trade-off assumption between relevance and fairness. Our code is available at https://github.com/Marfekey/TriRec.

Breaking User-Centric Agency: A Tri-Party Framework for Agent-Based Recommendation

TL;DR

This work proposes the first Tri-party LLM-agent Recommendation framework (TriRec) that explicitly coordinates user utility, item exposure, and platform-level fairness and finds that item self-promotion can simultaneously enhance fairness and effectiveness, challenging the conventional trade-off assumption between relevance and fairness.

Abstract

Recent advances in large language models (LLMs) have stimulated growing interest in agent-based recommender systems, enabling language-driven interaction and reasoning for more expressive preference modeling. However, most existing agentic approaches remain predominantly user-centric, treating items as passive entities and neglecting the interests of other critical stakeholders. This limitation exacerbates exposure concentration and long-tail under-representation, threatening long-term system sustainability. In this work, we identify this fundamental limitation and propose the first Tri-party LLM-agent Recommendation framework (TriRec) that explicitly coordinates user utility, item exposure, and platform-level fairness. The framework employs a two-stage architecture: Stage~1 empowers item agents with personalized self-promotion to improve matching quality and alleviate cold-start barriers, while Stage~2 uses a platform agent for sequential multi-objective re-ranking, balancing user relevance, item utility, and exposure fairness. Experiments on multiple benchmarks show consistent gains in accuracy, fairness, and item-level utility. Moreover, we find that item self-promotion can simultaneously enhance fairness and effectiveness, challenging the conventional trade-off assumption between relevance and fairness. Our code is available at https://github.com/Marfekey/TriRec.
Paper Structure (53 sections, 17 equations, 3 figures, 4 tables)

This paper contains 53 sections, 17 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. LLM-based Agentic Recommendation
  4. Creator-side Recommendation
  5. Platform Fair Ranking
  6. Problem Formulation
  7. Multi-Stakeholder Utility Modeling
  8. Agent-based Preference Interaction
  9. Method
  10. Generative Personalized Self-Promotion
  11. Platform-Agent Multi-Objective Re-Ranking
  12. Exposure as a Control State.
  13. State Representation.
  14. Action Space.
  15. State Transition.
  16. ...and 38 more sections

Figures (3)

  • Figure 1: Illustration of tri-party agentic recommendation and personalized item self-promotion.
  • Figure 2: Overview of the proposed two-stage TriRec framework, where Stage 1 performs relevance-aware re-ranking via user–item interaction, and Stage 2 conducts exposure-aware re-ranking through tri-party utility optimization.
  • Figure 3: Impact of varying the upper bound $\alpha_{\max}$ on user accuracy, platform fairness, and item utility on the CDs & Vinyl dataset. Moderate $\alpha_{\max}$ balances tri-party utilities.