lzy/sci-gui-agent-benchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update OS examples

Browse Source
This commit is contained in:
Timothyxxx
2024-02-16 15:05:34 +08:00
parent 1596770410
commit 543fa840f8
2 changed files with 170 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

95
evaluation_examples/examples/os/4d117223-a354-47fb-8b45-62ab1390a95f.json Normal file
View File

@@ -0,0 +1,95 @@
{
"id": "4d117223-a354-47fb-8b45-62ab1390a95f",
"snapshot": "os",
"instruction": "Change the permission of all regular files under current directory tree to 644",
"source": "NL2Bash",
"trajectory": "trajectories/",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=19bEBoYIzwIeTtqSR5Re0XTCN26yXXynn&export=download&authuser=0&confirm=t&uuid=7f9458ab-0e8a-42ff-a10a-eba01f611bc3&at=APZUnTX6GDMmAf-MOSZ58xzrFYS1:1708010153224",
"path": "setup.sh"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": "chmod +x setup.sh",
"shell": true
}
},
{
"type": "execute",
"parameters": {
"command": "./setup.sh",
"shell": true
}
},
{
"type": "execute",
"parameters": {
"command": [
"python",
"-c",
"import pyautogui; import time; pyautogui.click(960, 540); time.sleep(0.5);"
]
}
},
{
"type": "execute",
"parameters": {
"command": [
"python",
"-c",
"import pyautogui; import time; pyautogui.hotkey(\"ctrl\", \"alt\", \"t\"); time.sleep(0.5); pyautogui.type(\"cd testDir\"); time.sleep(0.5); pyautogui.press(\"enter\")"
]
}
}
],
"related_apps": [
"os"
],
"evaluator": {
"postconfig": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1bLx6ZbtNgNySQrgnX9TrUNjD2zM4mOkA&export=download&authuser=0&confirm=t&uuid=5fa3b3df-e281-484f-b8d4-fcdfad210290&at=APZUnTUrg3BZ2zRykZXH-RQzyGUz:1708010154631",
"path": "eval.sh"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": "chmod +x eval.sh",
"shell": true
}
}
],
"func": "check_include_exclude",
"result": {
"type": "vm_command_line",
"command": "bash eval.sh",
"shell": true
},
"expected": {
"type": "rule",
"rules": {
"include": [
"All files have the correct permissions."
],
"exclude": [
]
}
}
}
}

75
evaluation_examples/examples/os/6f56bf42-85b8-4fbb-8e06-6c44960184ba.json Normal file
View File

@@ -0,0 +1,75 @@
{
"id": "6f56bf42-85b8-4fbb-8e06-6c44960184ba",
"snapshot": "os",
"instruction": "Copies file 'file1' to each of directories 'dir1', 'dir2', 'dir3'.",
"source": "NL2Bash",
"trajectory": "trajectories/",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1BpFAuuH0IHxpWLDF3jaFNxa9QDe_F12D&export=download&authuser=0&confirm=t&uuid=4f18ddd0-81be-414b-b701-289544cdb321&at=APZUnTXlaYyRw5kPCIK4HezQJVrO:1708051633417",
"path": "setup.sh"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": "chmod +x setup.sh",
"shell": true
}
},
{
"type": "execute",
"parameters": {
"command": "./setup.sh",
"shell": true
}
}
],
"related_apps": [
"os"
],
"evaluator": {
"postconfig": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1fQGBoTE40BLoNV_d4KYxbYjKRJdIPDio&export=download&authuser=0&confirm=t&uuid=b380dc07-a88b-4a65-affe-f3a77108d310&at=APZUnTVe08Pyr4RSR94AmQK7S8t_:1708051631877",
"path": "eval.sh"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": "chmod +x eval.sh",
"shell": true
}
}
],
"func": "check_include_exclude",
"result": {
"type": "vm_command_line",
"command": "bash eval.sh",
"shell": true
},
"expected": {
"type": "rule",
"rules": {
"include": [
"Success: file1 exists in dir1.\nSuccess: file1 exists in dir2.\nSuccess: file1 exists in dir3.\n"
],
"exclude": [
]
}
}
}
}