codex-benchmark / summary.json
yzzhao's picture
Publish codex SIMPLE seed-0 evaluation evidence
24b126e verified
Raw History Blame Contribute Delete
5.83 kB
{
"planned_tasks": 84,
"published_results": 84,
"pending_tasks": 0,
"families": [
{
"id": "task01",
"name": "RoboCasa",
"total": 10,
"completed": 10,
"pending": 0,
"successes": 2,
"valid_results": 10,
"success_rate": 0.2,
"input_tokens": 63285354,
"cached_input_tokens": 62436736,
"output_tokens": 204733,
"usage_complete": 10,
"control_frequency_hz": 20,
"max_control_steps": 6000,
"preflight_results": 0,
"formal_results": 10,
"modified_results": 10,
"original_results": 0
},
{
"id": "task02",
"name": "LIBERO Long",
"total": 10,
"completed": 10,
"pending": 0,
"successes": 9,
"valid_results": 10,
"success_rate": 0.9,
"input_tokens": 12586499,
"cached_input_tokens": 12106752,
"output_tokens": 93969,
"usage_complete": 10,
"control_frequency_hz": 20,
"max_control_steps": 6000,
"preflight_results": 0,
"formal_results": 10,
"modified_results": 5,
"original_results": 5
},
{
"id": "task03",
"name": "RoboTwin",
"total": 10,
"completed": 10,
"pending": 0,
"successes": 7,
"valid_results": 10,
"success_rate": 0.7,
"input_tokens": 15621828,
"cached_input_tokens": 15138688,
"output_tokens": 105566,
"usage_complete": 10,
"control_frequency_hz": 25,
"max_control_steps": 7500,
"preflight_results": 0,
"formal_results": 10,
"modified_results": 7,
"original_results": 3,
"deferred_tasks": 0
},
{
"id": "task04",
"name": "RoboDojo",
"total": 42,
"completed": 42,
"pending": 0,
"successes": 31,
"valid_results": 42,
"success_rate": 0.7380952380952381,
"input_tokens": 259078099,
"cached_input_tokens": 254678912,
"output_tokens": 971405,
"usage_complete": 42,
"control_frequency_hz": 25,
"max_control_steps": 7500,
"preflight_results": 0,
"formal_results": 42,
"modified_results": 34,
"original_results": 8
},
{
"id": "task06",
"name": "SIMPLE G1",
"total": 12,
"completed": 12,
"pending": 0,
"successes": 7,
"valid_results": 12,
"success_rate": 0.5833333333333334,
"usage_complete": 12,
"control_frequency_hz": 50,
"max_control_steps": null,
"max_control_steps_by_task": {
"task06/01": 10000,
"task06/02": 15000,
"task06/03": 15000,
"task06/04": 15000,
"task06/05": 15000,
"task06/06": 15000,
"task06/07": 15000,
"task06/08": 10000,
"task06/09": 15000,
"task06/10": 15000,
"task06/11": 15000,
"task06/12": 15000
},
"preflight_results": 0,
"formal_results": 12,
"modified_results": 0,
"original_results": 12,
"input_tokens": 106763761,
"cached_input_tokens": 105331968,
"output_tokens": 303566
}
],
"interrupted_attempts": 4,
"usage_incomplete_tasks": [],
"execution_incomplete_tasks": [],
"deferred_tasks": [],
"new_evaluations": 12,
"progress": {
"finished": 84,
"running": 0,
"queued": 0,
"needs_review": 0,
"interrupted": 0,
"native_successes": 56,
"native_failures": 28
},
"simple_campaign": {
"planned": 12,
"published": 12,
"agent": "codex",
"model": "gpt-6-astra",
"effort": "high",
"seed": 0,
"automatic_episode_retries": 0
},
"simple_reruns": {
"authorized": 6,
"published": 6,
"remaining": 0,
"selection": "Latest validated completed result per task; earlier native failures remain in attempt history.",
"comparison_scope": "Only earlier failures were rerun at 15000 steps. Retained successes use 10000 steps. This selected-result view is not a uniform-budget agent comparison.",
"replacements": [
{
"task_key": "task06/02",
"current_job": "task06-02-transfer-between-tables-codex-seed0-attempt02",
"previous_job": "task06-02-transfer-between-tables-codex-seed0-attempt01",
"current_max_steps": 15000,
"previous_max_steps": 10000,
"success": false
},
{
"task_key": "task06/03",
"current_job": "task06-03-bend-pick-codex-seed0-attempt02",
"previous_job": "task06-03-bend-pick-codex-seed0-attempt01",
"current_max_steps": 15000,
"previous_max_steps": 10000,
"success": false
},
{
"task_key": "task06/04",
"current_job": "task06-04-bend-pick-and-place-codex-seed0-attempt02",
"previous_job": "task06-04-bend-pick-and-place-codex-seed0-attempt01",
"current_max_steps": 15000,
"previous_max_steps": 10000,
"success": true
},
{
"task_key": "task06/05",
"current_job": "task06-05-bend-handover-codex-seed0-attempt02",
"previous_job": "task06-05-bend-handover-codex-seed0-attempt01",
"current_max_steps": 15000,
"previous_max_steps": 10000,
"success": true
},
{
"task_key": "task06/06",
"current_job": "task06-06-handover-codex-seed0-attempt02",
"previous_job": "task06-06-handover-codex-seed0-attempt01",
"current_max_steps": 15000,
"previous_max_steps": 10000,
"success": false
},
{
"task_key": "task06/07",
"current_job": "task06-07-pick-and-place-and-hug-container-codex-seed0-attempt02",
"previous_job": "task06-07-pick-and-place-and-hug-container-codex-seed0-attempt01",
"current_max_steps": 15000,
"previous_max_steps": 10000,
"success": false
}
]
}
}