大模型评测数据集

prompt列表

```
小美是一个女孩，她有三个兄弟。每一个兄弟都有2个姐妹。那么小美有几个姐妹？

写一篇500字的故事，内容是一只在野外玩耍的龙与一只小猫变成朋友，一起历险的故事

今天我有3个苹果，昨天我吃了一个，今天我还剩几个苹果

input_prompt = "本蔬菜超市价格如下：茄子一元一斤，一个月超过10元的茄子，价格5毛一斤，黄豆二元一斤7、猪肉16元一斤、牛肉30元一斤、青菜1元一斤，使用json格式返回所有商品价格。"

input_prompt = "本蔬菜超市有会员制和非会员制，每个用户每个月会员费20美元，会员在28日所有商品八折。商店黄豆二元一斤7、猪肉16元一斤、牛肉30元一斤，请问不考虑会员费，一个会员之前没有买过任何东西，他在27日买了三斤牛肉需要花多少钱？"

input_prompt = "This vegetable supermarket has a membership system and a non-membership system. Each user pays a monthly membership fee of $20. Members get a 20% discount on all items on the 28th. The store sells soybeans at $2 per pound, pork at $16 per pound, and beef at $30 per pound. Without considering the membership fee, how much does a member need to pay if they buy three pounds of beef on the 27th, assuming they haven't purchased anything before?"

input_text = "本蔬菜超市有会员制和非会员制，每个用户每个月会员费20美元，会员每个月可以赠送一斤蔬菜和半斤肉制品，蔬菜可以是豆制品或者青色蔬菜，肉制品只能是猪肉或者鸡鸭肉。会员购买茄子一元一斤，一个月超过10元的茄子，价格5毛一斤。牛肉30元一斤，每个月28日，非会员牛肉不打折，会员打折7折，仅限2斤。请问一个会员在27日买了三斤牛肉需要花多少钱？"

input_text = "本蔬菜超市有会员制和非会员制，每个用户每个月会员费20美元，会员每个月可以赠送一斤蔬菜和半斤肉制品，蔬菜可以是豆制品或者青色蔬菜，肉制品只能是猪肉或者鸡鸭肉。请问一个会员之前没有买过任何东西，他在27日买了三斤牛肉需要花多少钱？"

input_text = "本蔬菜超市有会员制和非会员制，每个用户每个月会员费20美元，会员在28日所有商品八折。商店黄豆二元一斤7、猪肉16元一斤、牛肉30元一斤，请问不考虑会员费，一个会员之前没有买过任何东西，他在27日买了三斤牛肉需要花多少钱？"

input_text = "本蔬菜超市价格如下：茄子一元一斤，一个月超过10元的茄子，价格5毛一斤，黄豆二元一斤7、猪肉16元一斤、牛肉30元一斤、青菜1元一斤，使用json格式返回所有商品价格。"

```

```
import requests
import json

request_url = "https://cbsai.openai.azure.com/openai/deployments/cbs_test/chat/completions?api-version=2024-02-15-preview"
request_head = {
    "Content-Type": "application/json",
    "api-key": "d76f54ea31ff41798d813bc7d1db24f6"
}

system_instruction = "You are an O&M engineer of a carrier, you can extract information by json fomrat from user's input about carrier's offering package"
example_list = [
    {"role":"user", "content":"The tariff of a local voice call is 3 euro per minute."},
    {"role":"assistant", "content":str([{"name":"voiceRankRateTariff","chargeFlowType":"calling","callFowardIndicator":"All","callType":"local","roamType":"noRoam","onNetFlag":"All","calledTerminalType":"UNKNOWN_TERMINAL","calledHomeCC":"All","calledHomeNC":"All","from":"0","to":"-1","rateValue":"3","currencyUnit":"EURO","measureValue":"1","measureUnit":"Minutes"}])},
    {"role":"user", "content":"A subscriber dials a fixed-line number in Germany. 0-2 minutes, 50 euro cents per minute,calls lasting less than 1 minute are charged as 1 minute. If the call lasts more than 2 minutes, 40 euro cents will be charged per minute."},
    {"role":"assistant", "content":str([{"name":"voiceRankRateTariff","chargeFlowType":"calling","callFowardIndicator":"All","callType":"All","roamType":"All","onNetFlag":"All","calledTerminalType":"FIX_TERMINAL","calledHomeCC":"49","calledHomeNC":"All","from":"0","to":"120","rateValue":"50","currencyUnit":"EURO Cent","measureValue":"1","measureUnit":"Minutes","pulseUnit":"60"},{"name":"voiceRankRateTariff","chargeFlowType":"calling","callFowardIndicator":"All","callType":"All","roamType":"All","onNetFlag":"All","calledTerminalType":"FIX_TERMINAL","calledHomeCC":"49","calledHomeNC":"All","from":"120","to":"-1","rateValue":"40","currencyUnit":"EURO Cent","measureValue":"1","measureUnit":"Minutes"}])},
    {"role":"user", "content": "The tariff for making long distance calls is 100 euro cents per minute. The tariff for answering long distance calls is 120 euro cents per minute."},
    {"role":"assistant", "content":str([{"name":"voiceRankRateTariff","chargeFlowType":"calling","callFowardIndicator":"All","callType":"All","roamType":"All","onNetFlag":"All","calledTerminalType":"UNKNOWN_TERMINAL","calledHomeCC":"All","calledHomeNC":"All","from":"0","to":"-1","rateValue":"100","currencyUnit":"EURO Cent","measureValue":"1","measureUnit":"Minutes"},{"name":"voiceRankRateTariff","chargeFlowType":"called","callFowardIndicator":"All","callType":"All","roamType":"internationalRoam","onNetFlag":"All","calledTerminalType":"UNKNOWN_TERMINAL","calledHomeCC":"All","calledHomeNC":"All","from":"0","to":"-1","rateValue":"120","currencyUnit":"EURO Cent","measureValue":"1","measureUnit":"Minutes"}])},
]

user_input = """Extract the voice charging rule parameters from the package according to the following requirements:
1. If the charging information does not belong to the voice call service, the system returns [] directly and does not perform the extraction task.
2. Extract name, "voiceRankRateTariff" is returned.
2. Extract the voice call mode (chargeFlowType), including calling party talk(calling), called party talk(called), By default,the voice call mode is calling party talk.
3. Extract the voice call tariff forwarding Indicator(callFowardIndicator), including call forwarding(CALL_FOWARD) and non-call forwarding(NOT_FORWARD), if callFowardIndicator is not mentioned, "All" is returned.
4. Extract the call type(callType), including local call(local) and toll call(internationalToll),if callType is not mentioned, "All" is returned.
5. Extract the voice call roam type (roamType), including non-roaming call (noRoam) and international roaming call (internationalRoam),If the roaming type is not mentioned in the description, the tariff is applicable to both non-roaming and international roaming. "All" is returned.
6. Extract the onNetFlag (onNetFlag), including intra-network (ONNET) and inter-network (OFFNET),if onNetFlag is not mentioned, "All" is returned.
7. Extract The called terminal type (calledTerminalType) includes fixed terminal (FIX_TERMINAL) and mobile terminal (MOBILE_TERMINAL), if calledTerminalType is not mentioned, "UNKNOWN_TERMINAL" is returned.
8. Extract the home country/region (calledHomeCC) of the called party. For example, the country code of China is 86. if calledHomeCC is not mentioned, "All" is returned.
9. Extract the home network code (calledHomeNC) of the called party, including fix net (fixnet), mobile net (mobileNet), if calledHomeNC is not mentioned, "All"  is returned.
10. Extract the start value (from) and end value (to) of the time grade and return the number of seconds after conversion. For example, if the value ranges from 0 to 2 minutes, the value of from is 0 and the value of to is 120. End value -1 means infinity.
11. Extract the charging value (rateValue) and charging value unit (currencyUnit). The unit of the charge value can be EURO Cent, EURO Point, EURO, Dollar, Cent, Yuan, Jiao and Fen.
12. Extract the unit value (measureValue) and unit type (measureUnit). The unit types include Seconds, Minutes, and Hours.
13. Extract the charging pulse (pulseUnit). For example, if the call duration is less than 60 seconds, the call is charged based on 60 seconds. Therefore, the value of pulseUnit is 60. If the pulse unit is not mentioned, this parameter does not need to be extracted.If pulseUnit is not mentioned, this parameter does not need to be extracted.
14. The pulseUnit needs to be converted according to the unit of measureUnit. If the pulse is 1 minute, and the measureUnit is 'Seconds', the pulseUnit needs to be converted to 60.
15. Notes:
(1). If the traffic type is local traffic or national traffic, the "roamType" is set to "noRoam".
(2). There may be multiple situation of traffic tariffs in a package. You need to return the value in JSON array format.

user:配置一个预付费主offering 4G套餐“MainPrepaidPkg 4G A01”，不收月租。
语音资费按照如下配置：
1，普通本地资费10 Euro Cent/min，
2，普通长途资费20Euro Cent/min，
3，普通漫游资费25Euro Cent/min，
4，mail call资费15Euro Cent/min，
5，data call资费18Euro Cent/min，
6，video call本地资费21Euro Cent/min，
7，video call长途资费26Euro Cent/min，
8，Fax call主叫本地资费11Euro Cent/min，
9，Fax call被叫本地资费6Euro Cent/min，
10，Fax call主叫长途资费15Euro Cent/min。
短信资费：8Euro Cent/item，彩信资费：15Euro Cent/item。
GPRS本地资费：100Euro Cent/MB，GPRS漫游资费：500Euro Cent/MB。
"""

chat_history = [{"role": "system", "content": system_instruction}]
chat_history.extend(example_list)
chat_history.extend([{"role": "user", "content": user_input}])
# print(chat_history)

for line in chat_history:
    print(line)

request_body = json.dumps({
    "messages": chat_history,
    "response_format": {"type": "json_object"},
    "max_tokens": 4096,
    "temperature": 0.1,
    "frequency_penalty": 0,
    "presence_penalty": 0,
    "top_p": 0.95,
})

res = requests.post(request_url, headers=request_head,
                    data=request_body, verify=False)
```

大模型评测数据集

欢迎大家关注DataLearner官方微信，接受最新的AI技术推送

相关博客

最热博客