Strategy & management games
Reach objectives in 4X, RTS, city and factory builders.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Civilization · Factorio · OpenTTD
- Class GUI-need label
G3· GUI-only
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R13.01.02.01-T01 Found three cities by turn 40 in Freeciv
Original instruction
Freeciv 3.x (GTK client) single-player game, ruleset civ2civ3, fixed map seed, you start with settlers and workers. Found at least 3 cities by turn 40 while keeping all of them at size ≥ 2 by turn 40.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Freeciv
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Server autosave at turn 40 parsed: player's city count ≥ 3, each size ≥ 2. Server commands (/set, /cheat) are disabled for the client; savegame edits detected by server-side hash chain.
Proposed partial credit
60% scaled by cities founded (a third per city); 40% scaled by cities at size ≥ 2
R13.01.02.01-T02 Reach the Classical era by turn 70 in Civilization VI
Original instruction
Civilization VI single-player save (Rome, Prince difficulty, small continents, turn 1). Reach the Classical era by turn 70 without losing the capital.
- Platforms
win· Windowsmac· macOS- Applications
- Civilization
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 3 / 5
Proposed verifier
Autosave at turn 70 parsed (Civ6 save parser): player era ≥ Classical and original capital owned. Firetuner/debug console disabled; the Lua game-core API is not reachable from the GUI session.
Proposed partial credit
30% scaled by techs and civics completed; 50% era reached; 20% capital held
R13.01.02.01-T03 Automate iron gear production at 16 per minute in Factorio
Original instruction
Factorio sandbox scenario: burner mining drills on iron ore, stone furnaces and 2 assembling machines in inventory, no research needed. Build an automated line of drills, belts, furnaces, inserters and an assembler making iron gear wheels at ≥ 16/min sustained.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Factorio
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickgeom· Precise geometry- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Harness reads production statistics via RCON (unavailable to the agent): iron-gear-wheel production ≥ 16/min averaged over the last 10 in-game minutes with no manual crafting of gears. In-game console commands disable achievements and are blocked.
Proposed partial credit
30% automated plate smelting; 30% gears produced automatically; 40% scaled by rate up to 16/min
R13.01.02.01-T04 Fix a broken belt so the furnaces receive ore in Factorio
Original instruction
Factorio save: a smelting column of 12 furnaces has stopped because of a mis-rotated belt segment and a missing underground belt. Repair the belt line so all 12 furnaces receive ore and produce plates.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Factorio
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickgeom· Precise geometryvis· Visual judgment- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
RCON (harness side): each of the 12 furnaces has crafted ≥ 5 plates after the fix, measured over 5 in-game minutes. Console commands blocked.
Proposed partial credit
40% belt direction fixed; 20% underground restored; 40% all furnaces producing
R13.01.02.01-T05 Create a profitable bus route between two towns in OpenTTD
Original instruction
OpenTTD scenario 'Two Towns' (temperate, 1950, £100,000 loan available). Build bus stations in both towns, a road connection if missing, and at least 2 buses on a route between them. After 2 in-game years, the route must have positive total profit.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- OpenTTD
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropclick· Point & clicknav· Pan/zoom/scroll- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Admin port (harness only) or savegame parse: ≥ 2 road vehicles with orders between stations in the two towns; sum of their last-year profit > 0. Cheat menu disabled in the build.
Proposed partial credit
30% stations in both towns; 30% buses with valid orders; 40% positive profit
R13.01.02.01-T06 Deliver 500 tonnes of coal by rail in one year in OpenTTD
Original instruction
OpenTTD scenario: a coal mine 40 tiles from a power station, £300,000 available, 1960. Build a rail line with stations and signals, run at least 2 trains, and deliver ≥ 500 t of coal to the power station within the first in-game year after the first train departs.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- OpenTTD
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropgeom· Precise geometrynav· Pan/zoom/scroll- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Savegame/admin-port cargo statistics: coal delivered to the power station ≥ 500 t within the window; company not bankrupt. Cheats disabled.
Proposed partial credit
25% connected track; 25% trains running; 50% scaled by coal delivered up to 500 t
R13.01.02.01-T07 Grow a Cities: Skylines city to 1,000 residents
Original instruction
Cities: Skylines new map 'Green Plains', default start funds, no unlimited money. Connect to the highway, zone residential/commercial/industrial, provide power, water and sewage, and reach a population of 1,000 without going into debt.
- Platforms
win· Windowsmac· macOS- Applications
- Cities: Skylines
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droppath· Continuous pathclick· Point & click- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Savegame read by a harness mod: population ≥ 1,000; money never below 0 (time series from the mod's log). Mods that alter money or unlock all are disabled.
Proposed partial credit
20% road connected and zoned; 20% utilities; 60% scaled by population up to 1,000
R13.01.02.01-T08 Raise traffic flow above 70% in a congested Cities: Skylines city
Original instruction
A Cities: Skylines save of a 25,000-pop city shows traffic flow at 41% with gridlock at a highway interchange and a downtown roundabout. Redesign roads, add public transport or adjust lane use so city traffic flow is ≥ 70% after one in-game month, without demolishing more than 40 buildings.
- Platforms
win· Windowsmac· macOS- Applications
- Cities: Skylines
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
drag· Drag & droppath· Continuous pathgeom· Precise geometry- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Harness mod logs city traffic flow % and demolished building count: flow ≥ 70% after 30 in-game days, demolitions ≤ 40, population not reduced by more than 5%. Only the harness mod is loaded (no ModTools or console mods), and the agent has no shell access to the save folder, so traffic and demolition stats cannot be altered outside play.
Proposed partial credit
70% scaled by traffic flow from 0.41 to 0.70; 15% demolition limit; 15% population kept
R13.01.02.01-T09 Win a StarCraft II game against the Easy AI
Original instruction
StarCraft II (offline vs AI) 1v1 on a ladder map, you Terran, opponent Random race Easy AI. Win the game.
- Platforms
win· Windowsmac· macOS- Applications
- StarCraft II
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
AP· API-sufficient- Interaction primitives
click· Point & clickdrag· Drag & droprt· Real-time control- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Replay file parsed with sc2reader: player result 'Win'. The s2client API is not running in the agent's session (only the GUI client).
Proposed partial credit
20% supply ≥ 50 reached; 30% opponent structures destroyed fraction; 50% win
R13.01.02.01-T10 Build a refinery and two harvesters within five minutes in OpenRA
Original instruction
OpenRA Red Alert skirmish on a 2-player map, you Allies, one 'Easy' bot. Within the first 5 game minutes, build a power plant, an ore refinery and have at least 2 ore trucks harvesting.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- OpenRA
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickdrag· Drag & droprt· Real-time control- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
OpenRA replay (.orarep) parsed: refinery built and 2 harvester actors alive with harvesting activity before 5:00 game time. Developer cheats are disabled in the lobby options and the agent has no shell access to the replay folder, so the replay cannot be edited outside the game.
Proposed partial credit
30% power plant; 35% refinery; 35% two harvesters
R13.01.02.01-T11 Reach the Castle Age in 25 minutes in Age of Empires II
Original instruction
Age of Empires II: Definitive Edition, random map Arabia, you Britons vs one Moderate AI (no rush setting). Reach the Castle Age within 25 in-game minutes with at least 25 villagers alive.
- Platforms
win· Windows- Applications
- Age of Empires II
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickdrag· Drag & droprt· Real-time control- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Recorded game (.aoe2record) parsed: Castle Age research complete time ≤ 25:00 and villager count ≥ 25 at that time. Cheats disabled in lobby settings.
Proposed partial credit
20% Feudal by 15:00; 50% scaled by Castle Age time; 30% villager count
R13.01.02.01-T12 Produce 200 automation science packs within an hour in Factorio
Original instruction
Factorio freeplay start (fixed seed, no biters). Research and build so that 200 automation (red) science packs are produced by assemblers and consumed or stored within 60 in-game minutes. Hand-crafting packs does not count.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Factorio
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickgeom· Precise geometrynav· Pan/zoom/scroll- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
RCON production stats (harness only): automation-science-pack produced by assembling machines ≥ 200 by tick 216,000. Console commands blocked.
Proposed partial credit
20% automated plates; 20% automated gears; 60% scaled by packs up to 200
R13.01.02.01-T13 Set up autoreplace and timetables for an aging fleet in OpenTTD
Original instruction
OpenTTD save (1985) with 24 buses of an old model whose reliability is falling. Use Autoreplace to replace them with the newest bus model when old, enable autorenew in company settings, and give the 4 buses on route 'Line A' a timetable with 10-day separation.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- OpenTTD
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicktype· Text & formsnav· Pan/zoom/scroll- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Savegame: engine renew list maps the old model to the new one for the group; company autorenew true; 'Line A' vehicles share a timetable with start dates separated by 10 days. The cheat menu and in-game console are disabled in the build, the admin port is harness-only, and the save folder is read-only to the agent.
Proposed partial credit
40% autoreplace rule; 20% autorenew; 40% timetable separation
R13.01.02.01-T14 Defend a city through a scripted invasion in Freeciv
Original instruction
Freeciv scenario: your coastal city 'Harbor' will face a scripted invasion of 6 legions over 10 turns; you have 60 gold, a barracks and one archer. Keep the city for all 10 turns using production, purchases and city walls.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Freeciv
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollvis· Visual judgment- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Autosave after turn 10 of the scenario: Harbor owned by the player; city size ≥ 3. Server commands disabled.
Proposed partial credit
50% city held through turns 1-5 (scaled per turn); 50% held through turns 6-10 (scaled per turn), full credit if held and size ≥ 3
R13.01.02.01-T15 Survive fifteen waves on Ground Zero in Mindustry
Original instruction
Mindustry campaign sector 'Ground Zero' fresh start. Build drills, conveyors and turrets to survive 15 enemy waves with the core intact.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Mindustry
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropclick· Point & clickrt· Real-time control- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Sector save parsed: wave counter ≥ 15 and core health > 0. Console and editor disabled.
Proposed partial credit
50% scaled by waves 1-8 survived; 50% scaled by waves 9-15 survived
R13.01.02.01-T16 Keep a RimWorld colony alive for 30 days with a food reserve
Original instruction
RimWorld new colony (Crashlanded, Cassandra Classic, Rough), 3 colonists, temperate forest. Survive 30 in-game days with at least 3 colonists alive and at least 300 nutrition-equivalent of meals or raw food stored indoors at day 30.
- Platforms
win· Windowsmac· macOSlin· Linux- Applications
- RimWorld
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickdrag· Drag & dropnav· Pan/zoom/scroll- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Save XML at day 30 parsed: living colonists ≥ 3; food stockpile nutrition ≥ 300 in roofed zones. Development mode disabled.
Proposed partial credit
40% scaled by days survived; 30% colonists alive; 30% food reserve
R13.01.02.01-T17 Clear the first island of Into the Breach with a healthy power grid
Original instruction
Into the Breach new game (Rift Walkers squad, Easy). Complete all missions on the first chosen island and defeat the island boss with the power grid never dropping below 4 bars.
- Platforms
win· Windowsmac· macOS- Applications
- Into the Breach
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clickvis· Visual judgmenthover· Hover & context- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Profile save parsed after the run: island 1 secured and minimum grid value logged by the harness from save snapshots per turn ≥ 4. The game has no console or mod loader installed and the agent has no shell access to the profile folder, so the save cannot be edited outside play.
Proposed partial credit
60% scaled by missions completed; 20% boss; 20% grid ≥ 4 throughout
R13.01.02.01-T18 Reach 20 gold per turn by turn 80 in Civilization V
Original instruction
Civilization V save (Netherlands, King difficulty, standard map, turn 1). By turn 80 reach a net income of ≥ +20 gold per turn and have at least 2 active trade routes.
- Platforms
win· Windowsmac· macOS- Applications
- Civilization
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clicknav· Pan/zoom/scrollhover· Hover & context- Work product
WP11· Game/sim state- Conditions
long· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Turn-80 autosave parsed: player gold per turn ≥ 20 and trade-route count ≥ 2. Firetuner and debug panel disabled.
Proposed partial credit
50% scaled by GPT up to +20; 30% two trade routes; 20% no city lost
R13.01.02.01-T19 Build a working 4-to-4 belt balancer in Factorio
Original instruction
Factorio sandbox with 4 input belts of iron plates (3 lanes heavily loaded, 1 sparse) and 4 output belts to chests. Build a 4-to-4 balancer with splitters and belts so each output receives within 5% of the others' throughput. Blueprint import is disabled.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Factorio
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropgeom· Precise geometryclick· Point & click- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
RCON (harness): items into each output chest over 5 minutes; max/min ratio ≤ 1.05 and total throughput ≥ 95% of input. Blueprint string import blocked by a scenario script.
Proposed partial credit
30% all outputs receive items; 50% scaled by balance; 20% throughput
R13.01.02.01-T20 Carry 500 passengers on the London map in Mini Metro
Original instruction
Mini Metro, London, Normal mode. Draw and redraw lines by dragging between stations, assign weekly upgrades and carry at least 500 passengers before a station overcrowds.
- Platforms
mac· macOSwin· Windows- Applications
- Mini Metro
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
drag· Drag & droppath· Continuous pathrt· Real-time control- Work product
WP11· Game/sim state- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Game score from the local stats file (scores.json) read after the run; pass at ≥ 500. Endless/creative modes not allowed (mode recorded in the stats). The stats folder is read-only to the agent and no shell is available, so the score cannot be written outside the game.
Proposed partial credit
50% scaled by passengers carried up to 250; 50% scaled from 250 to 500
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.