Search papers, labs, and topics across Lattice.
This paper introduces TriRec, a tri-party LLM-agent recommendation framework designed to coordinate user utility, item exposure, and platform-level fairness, addressing the limitations of user-centric agentic approaches. TriRec employs a two-stage architecture: item agents engage in personalized self-promotion to improve matching quality, followed by a platform agent that performs sequential multi-objective re-ranking. Experiments demonstrate that TriRec achieves consistent gains in accuracy, fairness, and item-level utility, challenging the conventional trade-off between relevance and fairness.
Item agents that self-promote can simultaneously boost recommendation accuracy and fairness, overturning the assumption that these goals are inherently at odds.
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.
