Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation
The article introduces the Belief Engine (BE), a new framework designed for multi-agent deliberation using large language models (LLMs). BE aims to make stance dynamics more transparent by treating beliefs as evidential states and providing a structured memory for argument extraction. The framework allows for configurable updates based on evidence, enhancing the understanding of how agents change their stances during deliberation.
- ▪The Belief Engine provides an auditable layer for belief updates in multi-agent LLM interactions.
- ▪It treats beliefs as evidential states and exposes them as scalar stances.
- ▪Parameter sweeps across multiple LLMs show that BE can reliably shape stance dynamics while maintaining an evidence-level update trail.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.15343 (cs) [Submitted on 14 May 2026] Title:Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation Authors:Joshua C. Yang, Maurice Flechtner, Damian Dailisan, Michiel A. Bakker View a PDF of the paper titled Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation, by Joshua C. Yang and 3 other authors View PDF HTML (experimental) Abstract:LLM-based agents are increasingly used to simulate deliberative interactions such as negotiation, conflict resolution, and multi-turn opinion exchange.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
