An-Nego: LLM-based Strategic Anchoring Method for Multi-Issue Negotiations

ACL ARR 2026 January Submission7952 Authors

Keywords: LargeLanguageModels，Stakeholders， Multi-agent， Multi issue Negotiation

Abstract: LLM-powered negotiation agents must accurately identify and respond to other participants’ claims and interests to reach consensus. However, most prior work has focused on bilateral negotiation, leaving multi-party and multi-issue settings relatively underexplored, despite their prevalence in real-world semi-collaborative scenarios. We propose An-Nego, an anchor-based LLM-powered negotiation framework for multi-party stakeholder negotiation over multiple issues. An-Nego structures the process around anchored deal sets, participant voting, and moderator feedback to iteratively guide proposals and revisions. We evaluate An-Nego in terms of agreement success rate and time cost in multi-issue negotiation tasks. Experimental results show that An-Nego consistently outperforms baselines across multiple metrics, indicating improved agent capabilities in stakeholder negotiation environments. In addition, we further validate the effectiveness of the anchor effect in the multi-party gaming process.

Paper Type: Short

Research Area: AI/LLM Agents

Research Area Keywords: LargeLanguageModels， Stakeholders， Multi-agent， Multi issue Negotiation.

Languages Studied: English

Submission Number: 7952