Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation
Quick Answer
This paper shows that The Belief Engine (BE) introduces a configurable layer for multi-agent LLMs that audibly tracks belief updates during deliberation.
Quick Take
The Belief Engine (BE) introduces a configurable layer for LLMs that audibly tracks belief updates during deliberation. By utilizing a log-odds rule, BE effectively shapes stance dynamics based on evidence uptake and prior anchoring, outperforming traditional methods in reconstructing human deliberation outcomes on the DEBATE dataset.
Key Points
- Belief Engine audibly tracks belief updates in multi-agent LLM interactions.
- Utilizes log-odds rule to shape stance dynamics based on evidence uptake.
- Outperforms traditional methods on the DEBATE dataset for reconstructing opinions.
- Extracts arguments into structured memory for better stance management.
- Enables explicit study of evidence-grounded deliberation dynamics.
Paper Resources
📖 Reader Mode
~2 min readAbstract:LLM-based agents are increasingly used to simulate deliberative interactions such as negotiation, conflict resolution, and multi-turn opinion exchange. Yet generated transcripts often do not reveal why an agent's stance changes: movement may reflect evidence uptake, anchoring, role drift, echoing, or changed prompt and retrieval context. We introduce the Belief Engine (BE), an auditable belief-update layer that treats "belief" as an evidential state over a proposition and exposes it as scalar stance. BE extracts arguments into structured memory and updates stance with a log-odds rule controlled by evidence uptake u and prior anchoring a. Across multiple base LLMs, parameter sweeps show that these controls reliably shape stance dynamics while preserving an evidence-level update trail. On DEBATE, a human deliberation dataset with pre/post opinions, BE best reconstructs participants whose final stance follows extracted evidence; stable and evidence-opposed cases instead point to anchoring or factors outside the extracted evidence stream. BE provides configurable infrastructure for studying evidence-grounded deliberation, where openness, commitment, convergence, and disagreement can be tied to explicit update assumptions rather than hidden prompt effects.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2605.15343 [cs.AI] |
| (or arXiv:2605.15343v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2605.15343 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Joshua C. Yang [view email]
[v1]
Thu, 14 May 2026 19:13:12 UTC (2,052 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
Procedural Memory Distillation (PMD) enhances reinforcement learning by converting cross-episode signals into reusable memory, improving Qwen3-8B and OLMo3-Instruct-7B models by 3.8-5.5% on SCIKNOWEVAL and 7.9-13.6% on . The co-evolution of policy and memory allows for more effective self-supervision, demonstrating significant performance gains when both components are active.