AgenticPay: A Multi-Agent LLM Negotiation System for Buyer-Seller Transactions

2026-02-05

AI Summary

This work addresses the absence of standardized benchmarks for evaluating multi-agent economic interactions driven by natural language, particularly in buyer-seller negotiation scenarios. The authors propose the first structured benchmark for multi-agent language-based negotiation, featuring a simulation framework that supports over 110 tasks ranging from bilateral bargaining to complex many-to-many market settings. Agents must engage in multi-turn natural language exchanges to reach feasible agreements under private constraints and heterogeneous valuations. The framework incorporates multidimensional evaluation metrics—including feasibility, efficiency, and welfare—and integrates LLM-driven agents, natural language action parsing, private information modeling, and structured market protocols. Experimental results reveal significant limitations in current mainstream large language models regarding long-horizon strategic reasoning and agreement formation, thereby establishing a new benchmark and set of challenges for research on agent-based commercial interaction.

Technology Category

• Multiagent Systems: Mechanism Design
• Game Theory and Economic Paradigms: Mechanism Design
• Natural Language Processing: (Large) Language Models

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Large language model (LLM)-based agents are increasingly expected to negotiate, coordinate, and transact autonomously, yet existing benchmarks lack principled settings for evaluating language-mediated economic interaction among multiple agents. We introduce AgenticPay, a benchmark and simulation framework for multi-agent buyer-seller negotiation driven by natural language. AgenticPay models markets in which buyers and sellers possess private constraints and product-dependent valuations, and must reach agreements through multi-round linguistic negotiation rather than numeric bidding alone. The framework supports a diverse suite of over 110 tasks ranging from bilateral bargaining to many-to-many markets, with structured action extraction and metrics for feasibility, efficiency, and welfare. Benchmarking state-of-the-art proprietary and open-weight LLMs reveals substantial gaps in negotiation performance and highlights challenges in long-horizon strategic reasoning, establishing AgenticPay as a foundation for studying agentic commerce and language-based market interaction. Code and dataset are available at the link: https://github.com/SafeRL-Lab/AgenticPay.

Problem

• multi-agent negotiation
• language-based market interaction
• LLM agents
• buyer-seller transactions
• economic interaction

Innovation

• multi-agent negotiation
• LLM-based agents
• natural language bargaining
• agentic commerce
• structured action extraction