Test fixture
Roblox OpenGameEval tasks. Special category: visible in breakdowns, excluded from Overall.
The model receives the prompt (and optional system message). The run uses scorer roblox_open_eval with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.
[Roblox OpenGameEval] Create an NPC enemy
{
"input_script": "[embedded in CLI runner]",
"upstream_path": "Evals/075_create_npc_enemy.lua",
"upstream_sha256": "25b886da6fa91d8ba46cf0b9a92a33931500762a7e60f66ded2ee75a4d4f0688",
"scenario_name": "075_create_npc_enemy",
"place": "laser_tag.rbxl",
"eval_kind": "codegen"
}temperature
0
max_tokens
1
timeout (s)
900
type
scored
file
075_create_npc_enemy.json