chini-018-polling-station

Election Day Polling Station

One precinct, eight booths, three machines, a thousand voters, and the printer for ballot paper just jammed.

Source: Election administration literature, voter wait-time research, post-2020 polling reform reports

Prompt

Design the voter flow for a single polling precinct on a presidential election day.

Functional:
- Voter arrives, checks in at one of 4 poll-book stations (ID + signature). Issued a ballot.
- Voter takes ballot to one of 8 privacy booths to mark.
- Voter feeds marked ballot into one of 3 scanner/tabulator machines.
- Provisional ballots (registration mismatch) routed to a separate provisional table, sealed in envelope, NOT scanned.

Non-functional:
- A morning rush (4x baseline arrival) must NOT cause average wait to exceed 30 minutes. Booths/scanners must absorb.
- If a tabulator fails, ballots must be securely stored in the emergency-ballot bin for later scanning, NOT discarded or rerouted insecurely.
- If poll-book network goes down, check-in must continue via paper backup with reconciliation later. The line cannot stop.

Return a Chinilla CanvasState. Components: poll books, booths, tabulators, provisional table, emergency bin. Behaviors: queue (waiting line), split (regular vs provisional), ratelimit (booth turnover), circuitbreaker (network failover), storage (emergency bin).

Constraints

Max components: 13
Required behaviors: queue, split, circuitbreaker
Monthly budget: $8000

Stress scenarios

Steady turnout

baseline

Normal voter arrival, all systems up.

Pre-work rush

spike

4x arrival rate from 7-9am. Wait times must hold.

Tabulator down

outage

One scanner offline. Ballots must route to emergency bin, not discarded.

Poll-book network outage

outage

Check-in network down. Paper backup must keep the line moving.

Pass criteria (overall)

Min stability score: 65
Max drop rate: 5.0%
Min delivery rate: 92.0%
Max errors: 5

Submit your run

Submissions go through the chini-bench CLI. It calls your model with your key, scores the result locally, and posts to the leaderboard. Nothing leaves your machine except the canvas it produces.

End-to-end:

pip install git+https://github.com/collapseindex/chini-bench-cli.git
export OPENROUTER_API_KEY=...

chini-bench run chini-018-polling-station \
  --provider openrouter --model google/gemini-2.0-flash-001 \
  --as alice --x alice --linkedin alice-builds

Or inspect the prompt first:

chini-bench prompt chini-018-polling-station

Providers: openai · anthropic · google · openrouter · ollama

Leaderboard

Rank	Submitter	Model	Score	Stability	Delivery	Design	Pass	Links
#1	alex default	A anthropic/claude-sonnet-4.6	83	72.0	91.0	100.0	✗	X
#2	alex default	O openai/gpt-5.4	78	68.0	94.0	75.0	✗	X
#3	alex default	X x-ai/grok-4.20	74	70.0	63.0	100.0	✗	X
#4	alex default	G google/gemini-3.1-pro-preview	45	15.0	58.0	100.0	✗	X

Per-scenario breakdown of the top run

Scenario	Health	Drop rate	Delivered	Pass
baseline	74.0	1.5%	176	✓
morning-rush	72.0	1.7%	654	✓
tabulator-fail	66.0	1.4%	55	✗
pollbook-down	74.0	1.6%	161	✓

How is this scored? →