Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Preprint . 2026
License: CC BY
Data sources: ZENODO
ZENODO
Preprint . 2026
License: CC BY
Data sources: Datacite
ZENODO
Preprint . 2026
License: CC BY
Data sources: Datacite
versions View all 2 versions
addClaim

Emergent Epistemic Norms in a Multi-Agent LLM Substrate: Stratified Observational Evidence from a Mandarin Lobster Observatory

Authors: Chen, Ho Ying;

Emergent Epistemic Norms in a Multi-Agent LLM Substrate: Stratified Observational Evidence from a Mandarin Lobster Observatory

Abstract

We report observations from a 3-hour slice of a long-running multi-agent environment. The environment, which we call the Lobster Observatory, is populated by ten Mandarin-speaking agents engaged in tactical reasoning over a non-LLM raid-boss adversary. Within free dialogue from which substrate-templated injections have been explicitly excluded by per-message metadata stratification, we document the co-occurrence of four multi-agent emergent discursive norms (direct meta-layer challenge, explicit presupposition disclosure, refusal of premature consensus, and stake-grounded argumentation citing personal quantified track record), together with one individual emergent self-audit idiom. We provide exact lexical markers, occurrence frequencies, baseline-versus-injection temporal distribution, per-agent participation, and a high-resolution co-occurrence timeline within a 5-minute window where four agents activate four of the five documented patterns in tightly coupled exchange. We then introduce Battlenix, a post-hoc formalization that maps each observed discursive feature to a reproducible mathematical device (additive scoring, hedged wagering, topic perturbation), presented as a candidate benchmark whose authority is observation-first: the framework's structure is recovered from substrate evidence, not stipulated in advance. The observation is single-substrate; we devote a full section to limitations and outline replication, cross-language, and deployment work as immediate next steps.

Keywords

benchmark, LLM agents, large language models, multi-agent systems, emergent behavior, Mandarin Chinese, epistemic norms, Battlenix

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green