Skip to content

NSSL · Technion  /  Research project

Village Zero

Five language-model residents share a simulated village. Can they coordinate over time when each knows only part of what has happened?

An editorial engraving of the village square, with a tavern, paths, gardens and market stalls.
AI-generated illustration of the village. The simulated world has four connected locations.

What happens after a promise?

Suppose an innkeeper asks for supplies. The merchant can agree. For the tavern to serve an evening meal, someone must have stock, an offer must remain open, and the right resident must get an opportunity to collect it. Each step leaves evidence of its own.

Village Zero puts five decision makers in a persistent world so we can locate the point where a plan succeeds or breaks down. A chatbot conversation can end with the promise. A scripted scene can guarantee the meal. Here, each resident chooses at their own turn, and the next resident encounters the consequences of earlier choices.

That lets us ask a systems question alongside a behavioral one. When cooperation stalls, did the agent receive the relevant information, have a legal way to act and get a turn in time?

Sources · Tavern stage receipts · CSV  ·  Implementation references

Four places, five partial views

The square connects the tavern, market and residential area. Movement takes time. A visiting resident may see a peer or encounter an opportunity that someone elsewhere cannot see.

A day in the measured world

In Long-Life, a day contains two ticks. Walking from the square to the tavern or market takes one tick. The route to the homes takes two. A journey can continue on the clock after the resident’s initiating decision.

Roles change what a resident can do locally. Talia can pack produce at the homes when garden reserves, stock and the day’s work allow it. A local sale requires stock, a co-located buyer and an available offer. Residents can also deliver messages, invite one another to rest and decline.

Checkpoint R06 · end of day 24

Who was where?

Recorded positions
LocationResidents at tick 48
TavernMara, innkeeper · Nessa, village gossip
MarketIlan, merchant · Toren, smith
HomesTalia, gardener
Village squareNo resident present

Sources · R06 checkpoint and route graph  ·  Implementation references

Who gets to change the world?

Five residents share a language model. On Toren’s turn it sees the smith’s surroundings, selected memories and the actions open to him. A promise to pay can reach a neighbour as speech. Moving two coins requires him to choose a valid purchase and the engine to apply it to both accounts.

Giving the World Engine custody of the accounts makes a sale independently checkable. Limiting each resident’s view lets us trace how news travels. Giving the scheduler a fixed number of turns makes attention a resource that residents can lose while an offer is open.

Sources · Implementation references  ·  R07/B · memory selection for call 294

Why did Nessa get to answer?

In R06/A two things had to line up. Nessa needed a turn, and Mara’s invitation had to be actionable when it arrived. The records give a different origin for each.

Event 100 at tick 51 provides memory 151 and scheduler job 79. Separately, Mara opens offer 101 at tick 52. Nessa’s scheduled turn at tick 55 includes a legal join action. She selects it and event 105 records the shared rest.
The scheduler selected a job created before Mara issued the invitation. The live offer supplied a choice inside Nessa’s turn. Receipt 105 links her choice to offer 101.
Inspect the steps and their record identifiers
  1. Where the turn came fromAt tick 51, free-time event 100 creates observation 151 in Nessa’s notebook. That observation supplies the trigger for scheduler job 79.
  2. Where the invitation came fromAt tick 52, Mara chooses to invite Nessa. The engine opens offer 101 in tavern encounter 7, with expiry recorded as tick 57. The offer requests no new scheduler job.
  3. What the scheduler selectedJob 79 receives a slot at tick 55. Nessa’s call 55 still identifies the earlier spatial opportunity as its trigger.
  4. What was legal on that turnIn a later tick and still together in the tavern, the two residents meet the conditions for a response to offer 101. The packet lists joining, declining and resting alone.
  5. What Nessa selectedHer reply chooses rest:join:innkeeper. The engine checks the live offer, encounter and participants before applying the operation.
  6. What the engine recordedEvent 105 links the shared rest to offer 101 and records its end at tick 56. These records establish one joint activity.

Sources · R06/A · job 79, offer 101 and receipt 105  ·  Implementation references  ·  Open the full-size decision diagram

Research context

Imagine inviting a neighbour to a party. Smallville asks whether invitations travel and guests arrive. Concordia asks how a scenario should interpret a participant’s intended action. AI Town makes conversations and movement part of a persistent town that people can join.

Village Zero asks a narrower question about that same invitation. Did the neighbour ever receive a turn while accepting was possible, and what did the engine record? Keeping the scheduler, the menu and the outcome separately observable lets us investigate a missed meeting.

Primary sources · Park et al. on Smallville  ·  DeepMind on Concordia’s Game Master  ·  AI Town project

Compare worlds with a common past

A branch starts from a saved state. Before reading a performance difference as behavior, we ask what options the two branches exposed and how much model time each received.

Pair the histories

In Long-Life, eight separately seeded histories ran for 24 days. We copied each tick-48 checkpoint into A and B, assigned the same generation seed within a pair, then observed the next 72 days.

Keep the units visible

An invitation, a prompt that offers a response and a completed joint rest are three observations. We compare branches within each history and report the paired roots. Repeated turns share a world.

Replay a narrow question

A saved state can isolate a particular offer. We can inspect the available actions, vary the buyer’s next turn and use a scripted legal purchase as a mechanical control.

Sources · Eight paired histories · CSV  ·  R06/A · job 79, offer 101 and receipt 105  ·  Six replay outcomes · CSV

More turns change some outcomes more than others

B delivered more messages in all eight matched histories. Completed joint rests rose in three, fell in two and tied in three. The median paired change in rests was zero.

For R01 to R08, B minus A message counts are 23, 13, 111, 234, 218, 49, 48 and 84. Joint-rest differences are 12, 1, 4, 0, −5, −8, 0 and 0.
Each mark compares B with A from the same checkpoint over 72 simulated days. The panels use different scales. Zero means the two branches recorded the same count.

Was joining an option?

A recorded 32 completed joint rests across 38 turns that offered a join. B recorded 36 across 74 such turns. In R06 the opportunity appeared nine times in A and once in B. Every one produced a joint rest.

What did the extra turns cost?

Across the eight 72-day continuations, A used 1,152 model calls and 148.9 minutes of model evaluation. B used 2,943 calls and 387.6 minutes. The two schedules exposed different residents and opportunities at different costs.

Sources · Eight paired histories · CSV  ·  112 turns offering joint rest · CSV  ·  Runtime by phase · CSV

Information · #2

What reached the merchant?

In an earlier three-resident village, all 42 scored detail instances from six direct accounts reached the merchant’s input. Six were fully present in the next utterance. See the three routes.

Cooperation · #21

Did the tavern serve its guests?

Seven of 12 controlled three-resident worlds completed the meal service. Four autonomous starts produced four prepared crates and two collections. None reached service within the observation window. Follow the crates.

Timing · #72–73

When did the buyer act?

From one saved offer, both immediate-turn replays produced a purchase. All four deferred replays reached expiry without one. A follow-up gave a buyer who inspected an offer another chance to act. Compare the two interventions.

Sources · Telephone Game scoring · CSV  ·  Tavern stage receipts · CSV  ·  Six replay outcomes · CSV  ·  Experiment 73 · 24 continuations and their outcomes  ·  Study designs, models and observation windows

A village on one GPU

Including the 24-day common histories, the Long-Life run made 4,479 decision calls to qwen3.8:27b on an NVIDIA RTX A5000. Processing the prompts and generating the replies took 9.76 model hours. The complete job, through archiving, took about 11 hours 13 minutes.

The model processed 17,230,730 input tokens and generated 768,521 output tokens. Restricting what enters a turn therefore matters both for what the resident knows and for how much computation the experiment takes.

What would count as progress?

The current system can settle a sale and record shared activity. Sustaining cooperation across many opportunities is a larger empirical question.

Long-term autonomy

Repeat quiet and active histories from more starting states. Count turns with feasible actions, the length of quiet intervals and which events end them.

Memory and social change

A retrieved record can appear before a decision. A controlled removal or replacement, with the same legal options and scheduler, could test whether it changes what the resident does.

Useful behavior per unit of compute

Compare schedules at matched call budgets as well as matched simulated time. Measure completed activities, who received a turn and how opportunities were distributed.

Sources · Eight paired histories · CSV  ·  52 turns with a purchase option · CSV  ·  Six replay outcomes · CSV

Read the research journal

Begin with the five residents. Then follow a single crate through conversation, a legal offer and replay.

Project essay

A World in Five Views

Five residents can stand in the same village and have different evidence about what happened. Follow a parcel and a tavern invitation to see how the experiment works.

Experimental case study

One Crate, Two Coins

Toren has four coins. Ilan offers him a crate for two. Follow the purchase to its receipt, then return to the open offer and change when the smith gets his turn.