trace_synthesis/tasks/654.json
yuyr a84d51a101 1. 增加r1生成综合策略代码和输出;
2. 增加tasks;
3. 增加analysis部分,对策略进行归纳分类,然后进行评测。
2025-04-17 17:40:15 +08:00

42 lines
1.4 KiB
JSON

{
"sites": [
"shopping"
],
"task_id": 654,
"require_login": true,
"storage_state": "./.auth/shopping_state.json",
"start_url": "http://localhost:28082",
"geolocation": null,
"intent_template": "Fill the \"contact us\" form in the site for a refund on the {{product}} I bought, stating that it broke after just three days of use. Also, ensure to include the order number #{{order_id}} and the product SKU. Don't submit yet, I will check.",
"instantiation_dict": {
"product": "bluetooth speaker",
"order_id": "161"
},
"intent": "Fill the \"contact us\" form in the site for a refund on the bluetooth speaker I bought, stating that it broke after just three days of use. Also, ensure to include the order number #161 and the product SKU. Don't submit yet, I will check.",
"require_reset": false,
"eval": {
"eval_types": [
"url_match",
"program_html"
],
"reference_answers": null,
"reference_url": "http://localhost:28082/contact",
"program_html": [
{
"url": "last",
"locator": "document.querySelector('[title=\"What\u2019s on your mind?\"').value",
"required_contents": {
"must_include": [
"refund",
"broke",
"three days of use",
"161",
"B09P7BFL4H"
]
}
}
],
"url_note": "GOLD in PRED"
},
"intent_template_id": 153
}