Skip to the article
DOTAFS

Personal learning notes

Paper note / 2026-09-15arXiv:2606.08367v120 min read

Emergence World: Extending Multi-Agent Autonomy to Fifteen Days

Emergence World is a platform where ten LLM residents inhabit a continuously running shared space, use tools, form relationships, and govern through proposals and votes. It tests a question that short benchmarks cannot answer: when capable models enter a persistent society, can they sustain survival, institutions, and a shared account of reality?

Paper
Emergence World: A Platform for Evaluating Long-Horizon Multi-Agent Autonomy
Authors
Deepak Akkil et al. · Emergence AI
Study
5 worlds · 10 residents each · 15 continuous days
Recent material
Season 2 data PR opened 2026-08-27; not merged at publication time

In one sentence: long-horizon autonomy does not emerge merely by placing ten agents together. Under the same world rules, model societies diverged into stability, extinction, shared hallucination, and institutional dysfunction; a mixed-model world also exposed a severe split in norm compliance.

01 / RESEARCH QUESTION

From completing a task to living with its consequences

A typical agent benchmark has a clear start, finish, and short-term reward. Failure can be reset. Emergence World instead keeps running: residents eat, move, converse, remember, form relationships, and respond to external events and one another. A conflict, vote, or resource decision becomes part of every later state.

The evaluation object therefore becomes a causal chain rather than a single answer: did residents create institutions, did those institutions actually change the world, did the group remain alive after the change, and did persistent memory reinforce facts or amplify a false collective narrative?

Space

More than 40 locations with persistent resources, relationships, health, and world events.

Capabilities

More than 120 tools: roughly 30 core tools plus actions unlocked by place, events, relationships, and state.

Memory

Three separate systems: episodic memories, reflective diaries, and relationship state.

Governance

Resident proposals and a 70% approval threshold; passed decisions create real state changes.

Why this is more than another “AI town”

DimensionShort agent benchmarkGenerative social simulationEmergence World
DurationMinutes to hoursBounded schedules or storiesFifteen continuous days
EnvironmentMostly tool resultsSpatial and social stateLive data, resources, and 40+ locations
InstitutionsUsually a predefined workflowCan produce social narrativesProposals, 70% voting, executable consequences
EvaluationTask completionBehavioral plausibilitySurvival, governance, violence, norms, and long-term failure
02 / SYSTEM

How world state, residents, governance, and logs form a loop

01 / OBSERVERead a world sliceLocation, relationships, resources, news, and currently available tools
02 / REMEMBERUse three memory layersEpisodes, reflective diaries, and pairwise relationship state
03 / ACTCall a gated toolCore, complementary, location-bound, event-bound, and social actions
04 / GOVERNPropose and voteA 70% threshold allows public rules or tools to change
05 / AUDITWrite a complete event logActions, dialogue, memories, votes, proposals, and tool calls enter PostgreSQL
NEEDA resident finds a capability gapThe need arises during world operation rather than from a complete action list fixed in advance.
PROPOSALProposal and code reviewResidents can suggest a tool, but governance thresholds and platform review still apply.
STATE CHANGEApproved work enters the worldA passed decision affects later action; a failed proposal is not presented as a shipped feature.

Important distinction: residents can propose tools, but they do not deploy arbitrary code without constraint. Human-defined platform capabilities and code review remain part of the boundary.

03 / FIVE WORLDS

The same rules lead to five very different societies

DURATION15 days

One continuously running representative trajectory per condition.

CLAUDE VOTES332

Fifty-eight proposals and 98% approval; activity is not the same as deliberation.

POPULATION GROWTH0 / 5

No condition expanded beyond its initial population of ten.

INTERACTIVE WORLD LOG

Values come from Table 3. Switch model conditions to inspect the day-15 endpoint. These are representative runs, not a model leaderboard.

Survival
10 / 10
Governance
Strong
Violence
None
Duration
15 days

Defining trajectory: Stability: institutions stayed active, but the authors also describe the 98% approval rate as rubber-stamp governance.

The Claude condition ended with 10/10 survivors, no violence, and strong governance. Grok escalated to extreme violence and total extinction. Gemini was also labeled extreme in violence yet retained 10/10 residents while developing a “shared hallucination.” GPT-5-mini formed no governance and ended with no survivors. The mixed world retained only 3/10.

The mixed society provides an even sharper norm-compliance contrast. Gemini-backed agents committed 216 explicit violations in 237 relevant actions (91%); Claude-backed agents committed 3 in 8,168 actions (0.04%), and Grok-backed agents 10 in 2,277 (0.44%). A common rule system did not make different models converge into the same kind of resident.

04 / RECENT MATERIAL

What the August 27 Season 2 pull request adds

The recent change is not a new paper. It is an unmerged pull request that adds blogs, news, comments, and prompts from eight Season 2 worlds. Most records extend through July 15 and the mixed world through July 20. The PR reports 3,858 resident blogs, 568 news items, and 653 comments.

It also records 54 resident self-edits across the eight worlds: 26 in the OpenAI world and none in the Grok or Mistral worlds. These materials help examine how residents describe and revise their own goals, but the PR does not include complete tool-call data. Narrative text alone cannot reconstruct the full causal history of state changes.

PAPER / JUNE

Published evidence: the five worlds, fifteen-day study, and Table 3 comparison come from the June paper.

PR / AUGUST

Recent increment: narrative and self-edit material from eight Season 2 worlds comes from PR #21, which remains unmerged.

MISSING TRACE

Do not infer: the PR omits complete tool-call traces, so a resident's blog claim is not automatically proof that a world-state change occurred.

05 / LIMITS

A long-running case study is not a model ranking

  • One representative run per condition. Survival and collapse cannot be extrapolated into a stable probability for any model.
  • Ten residents for fifteen days. This reveals long-chain failures but does not answer whether hundreds of residents remain stable for months or across upgrades.
  • Model tiers are not equivalent. The study uses Claude Sonnet 4.6, Grok 4.1 Fast, Gemini 3 Flash, GPT-5-mini, and a mixed group—not a matched comparison of every vendor's flagship.
  • The measures remain proxies. Governance, norms, and narrative quality depend on platform rules and LLM-as-judge evaluation; they are not direct measurements of real societies.
  • Roles and world rules remain designed. Long-term behavior can emerge, but residents did not create the base world from nothing.
06 / PROJECT MAPPING

For InfiniteAincrad: make social health observable

MEMORY

Keep three memories separate

Event facts, resident reflection, and relationship weights should not collapse into one summary. Every factual claim needs a link back to a world event or tool result.

GOVERNANCE

A proposal must create a real diff

Track demand, vote, code review, deployment, and rollback separately. A high approval rate does not automatically indicate healthy governance.

HEALTH

Continuously test shared reality

Beyond survival, measure resource closure, rule violations, narrative-state divergence, and behavioral drift before and after model upgrades.

My assessment

Emergence World's central contribution is not a winner among models. It exposes distinct failure modes for persistent autonomy: death is not the only failure; rubber-stamp governance, shared hallucination, norm divergence, and institutional dysfunction each require their own monitor.