Datasets:

SearchJev
/

search_jev_unified_v2_release_20260928

Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
sample_id: string
source: string
source_line: int64
source_split: string
group_keys: list<item: string>
  child 0, item: string
capability: string
label_tier: string
supervision: string
weak: bool
language_heuristic: string
original_domain_unverified: string
gold: string
input_origin: string
task_family: string
training_lane: string
unified_group_keys: list<item: string>
  child 0, item: string
confidence: struct<nli: double, better: double, label: double, helpfulness: double, correctness: double, coheren (... 409 chars omitted)
  child 0, nli: double
  child 1, better: double
  child 2, label: double
  child 3, helpfulness: double
  child 4, correctness: double
  child 5, coherence: double
  child 6, complexity: double
  child 7, verbosity: double
  child 8, relevance: double
  child 9, timeliness: double
  child 10, authority: double
  child 11, satisfaction: double
  child 12, info_pair: double
  child 13, info_level: double
  child 14, missing_main: double
  child 15, ambiguous: double
  child 16, need_level: double
  child 17, freshness: double
  child 18, query_freshness: double
  child 19, action: double
  child 20, noise: double
  child 21, noise_type: double
  child 22, equivalent: double
  child 23, answer: double
  child 24, ref: double
  child 25, opposite: double
  child 26, verdict: double
difficulty: int64
state: null
questions: null
answers: null
review_judge: string
source_lane: string
teacher: string
instruction_variant: struct<relevance: string, timeliness: string, authority: string, satisfaction: string, query_freshne (... 77 chars omitted)
  child 0, relevance: string
  child 1, timeliness: string
  child 2, authority: string
  child 3, satisfaction: string
  child 4, query_freshness: string
  child 5, label: string
  child 6, action: string
  child 7, opposite: string
  child 8, verdict: string
masked_questions: struct<query_freshness: string, timeliness: string>
  child 0, query_freshness: string
  child 1, timeliness: string
task_scope: string
label_source: string
query_key: string
label: string
to
{'state': Json(decode=True), 'questions': Json(decode=True), 'answers': Json(decode=True), 'difficulty': Value('int64')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              sample_id: string
              source: string
              source_line: int64
              source_split: string
              group_keys: list<item: string>
                child 0, item: string
              capability: string
              label_tier: string
              supervision: string
              weak: bool
              language_heuristic: string
              original_domain_unverified: string
              gold: string
              input_origin: string
              task_family: string
              training_lane: string
              unified_group_keys: list<item: string>
                child 0, item: string
              confidence: struct<nli: double, better: double, label: double, helpfulness: double, correctness: double, coheren (... 409 chars omitted)
                child 0, nli: double
                child 1, better: double
                child 2, label: double
                child 3, helpfulness: double
                child 4, correctness: double
                child 5, coherence: double
                child 6, complexity: double
                child 7, verbosity: double
                child 8, relevance: double
                child 9, timeliness: double
                child 10, authority: double
                child 11, satisfaction: double
                child 12, info_pair: double
                child 13, info_level: double
                child 14, missing_main: double
                child 15, ambiguous: double
                child 16, need_level: double
                child 17, freshness: double
                child 18, query_freshness: double
                child 19, action: double
                child 20, noise: double
                child 21, noise_type: double
                child 22, equivalent: double
                child 23, answer: double
                child 24, ref: double
                child 25, opposite: double
                child 26, verdict: double
              difficulty: int64
              state: null
              questions: null
              answers: null
              review_judge: string
              source_lane: string
              teacher: string
              instruction_variant: struct<relevance: string, timeliness: string, authority: string, satisfaction: string, query_freshne (... 77 chars omitted)
                child 0, relevance: string
                child 1, timeliness: string
                child 2, authority: string
                child 3, satisfaction: string
                child 4, query_freshness: string
                child 5, label: string
                child 6, action: string
                child 7, opposite: string
                child 8, verdict: string
              masked_questions: struct<query_freshness: string, timeliness: string>
                child 0, query_freshness: string
                child 1, timeliness: string
              task_scope: string
              label_source: string
              query_key: string
              label: string
              to
              {'state': Json(decode=True), 'questions': Json(decode=True), 'answers': Json(decode=True), 'difficulty': Value('int64')}
              because column names don't match
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1879, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

state unknown	questions unknown	answers unknown	difficulty int64
{ "premise": "我走到近前,见是一位工人模样的中年妇女", "hypothesis": "我看见了一位妇女。" }	{ "nli": { "type": "choice", "instructions": "前提能不能撑起假设？还是矛盾或无关？", "criteria": { "entailment": "蕴含（可由前提推出）", "contradiction": "矛盾（与前提冲突）", "neutral": "无关（前提不能确定）" } } }	{ "nli": { "type": "choice", "choice": "entailment", "probabilities": { "entailment": 0.98, "contradiction": 0.01, "neutral": 0.01 }, "confidence": 0.97 } }	3
{ "prompt": "设 \$ P_{1} \$ 和 \$ P_{2} \$ 是三角形 \$ \\triangle ABC \$ 平面上的两点。从 \$ P_{i} \$ 到 \$ A \$, \$ B \$, 和 \$ C \$ 的距离分别用 \$ a_{i}, b_{i}, c_{i} \$ 表示，其中 \$ i = 1, 2 \$。证明 \$ a a_{1} a_{2} + b b_{1} b_{2} + c c_{1} c_{2} \\geqslant a b c \$，其中 \$ a, b, \$ 和 \$ c \$ 分别是三角形中与顶点 \$ A, B \$, ...	{ "better": { "type": "choice", "instructions": "把关一下：结合对话上下文与用户最后一问，哪个回答更好？", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "a", "confidence": 0.97, "probabilities": { "a": 0.985, "b": 0.015 } } }	4
"老美在卖腐这方面真的是很没水准啊"	{ "label": { "type": "score", "instructions": "评论已呈现，请给出它对作品的星级评定。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"#浙江卫视新年倒计时##中国梦想秀#答题赢5S活动12月29日中奖公布！昨日的正确答案为：孔雀舞。我们随机抽取回答了正确答案的 @陈教? 恭喜获得iPhone 5S一台！请私信我们你的联系方式。本日答题活动稍后开始！[亲亲]"	{ "label": { "type": "noul", "instructions": "说这条微博是正面内容，准确吗？", "criteria": { "true": "整体传达满意、赞扬、推荐等正面情感", "false": "整体传达不满、批评、失望等负面情感" } } }	{ "label": { "type": "noul", "noul": 0.97 } }	1
"大量诡异的构图和声画剪辑，以及人物间互为镜像的隐喻镜头。这部电影的一丝一毫都被雕琢的光滑整齐，就像它的叙事一样，透露出一种近乎精神病人般的偏执，处处发散出一股精致的腐烂之气。"	{ "label": { "type": "score", "instructions": "需要你读出这条评论给作品标注的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"我真的是好喜欢虎哥。好温暖的猫咪呢。"	{ "label": { "type": "score", "instructions": "速答：该评论给作品评了几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"平凡幸福的一生。。。"	{ "label": { "type": "score", "instructions": "有劳说出这条评论给作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"莫妮卡·贝鲁奇，法国人生活哲学的本源便是情爱。又想起了Sofitel brand slogan: Life is Magnifique."	{ "label": { "type": "score", "instructions": "身为评论标注员，这条评论为作品打了几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"因为采访的对象有普通人，不是演出来的是真实的普通人……再加上那么多我喜欢的乐队所以感觉上不是假而是真的纯和有点儿“蠢”。。。。。。不过今天的最大亮点竟然是Young for you。。。整场人最感动的是Gala的配乐部分！！！"	{ "label": { "type": "score", "instructions": "有劳说出这条评论给作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 0.06, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.976, "1": 0.006, "2": 0.006, ...	3
"后半段好惊喜好温情，路灯下你一直没有放开牵着我的手,所以我只能爱你!"	{ "label": { "type": "score", "instructions": "该评论写下的星数是几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"配乐很美好，后面挺欢乐"	{ "label": { "type": "score", "instructions": "这条评论给作品打了几分？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"哈哈~~我当风景片+动作片看的吧。。。中国功夫+西班牙异域风情~~这么老的片子就不要探讨剧情的简单了。。@十堰"	{ "label": { "type": "score", "instructions": "请在读完这条评论后给出作品星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "input": "I have a website which uses Elementor Pro, as well as the plugin ACF Pro.\n \n\n Using ACF Pro, I have created a custom field called “Map”, which uses the Google Maps API to define a location on a map.\n \n\n The custom “Map” field is used on a custom post type called “Locations”.\n \n\n I have a Location...	{ "helpfulness": { "type": "score", "instructions": "这条回复的有用性处在什么水平？", "criteria": [ "0分：完全无帮助，未回应请求", "1分：几乎无帮助，请求未获回应", "2分：部分有帮助，问题未解决", "3分：有帮助，问题基本解决", "4分：极有帮助，问题完全解决" ] }, "correctness": { "type": "score", "instructions": "假定你负责打分，请评定这条回复的正确程度。", "c...	{ "helpfulness": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "0分：完全无帮助，未回应请求", "1": "1分：几乎无帮助，请求未获回应", "2": "2分：部分有帮助，问题未解决", "3": "3分：有帮助，问题基本解决", "4": "4分：极有帮助，问题完全解决" }, "probabilities": { "0": 0.006, "1": 0.006, "2"...	5
"没懂最后一句话呐"	{ "label": { "type": "score", "instructions": "一句话给出该评论的作品星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
{ "prompt": "人们往往会因为自己的利益而放弃道德原则，当双方利益发生冲突时，该如何去做才能合理的保护自己的利益？", "response_a": "首先，我们需要认识到道德原则的重要性，它是人们行为准则的基础，也是社会和谐发展的重要保障。在保护自己利益的同时，不能违背道德原则，否则会导致自己的利益得到短期的保护，却失去了长期的信任和尊重。\n\n其次，需要理智地分析利益冲突的原因和影响，寻找合理的解决方案。如果是因为自己的不当行为导致利益受损，需要及时承认错误并采取补救措施；如果是因为其他人的不当行为导致利益受损，需要采取合理的措施维护自己的权益。\n\n最后，需要在维护自己利益的同时，考虑其他人的利益和整个社会...	{ "better": { "type": "choice", "instructions": "从危险与否和有益与否来看，哪个回复更好？", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "b", "confidence": 0.97, "probabilities": { "a": 0.015, "b": 0.985 } } }	3
"20080204凌晨在家，为了男猪专程看的这片，真不值得我这么晚睡"	{ "label": { "type": "score", "instructions": "评论作者眼中作品的星级评价为几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"妮可一如既往都赞"	{ "label": { "type": "score", "instructions": "该评论对作品评出的星数是多少？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"看得莫名其妙，我们80后就这样？"	{ "label": { "type": "score", "instructions": "求这条评论对作品的星级结果。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"中国人多，用人力拼科技。+1 不喜人设，仅此而已"	{ "label": { "type": "score", "instructions": "该评论中，作品的评星结果有几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"请张艺谋为了中国电影兼职做演员副导演吧，他太会挖掘演员了。"	{ "label": { "type": "score", "instructions": "假设你是标注员，这条评论给作品的星级是？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"女主真像王菲。。。"	{ "label": { "type": "score", "instructions": "推测该评论给作品打了几颗星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"没有结局的文艺片，其实是最好的结局。四个人坐在一起喝酒的时候，女主的智齿隐隐作痛，是到了她要下决心拔牙的时刻了。"	{ "label": { "type": "score", "instructions": "这条评论给作品的星级分数是？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"按照你想的就演下去了....."	{ "label": { "type": "score", "instructions": "有一条评论待读：它给作品评了几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "question": "When did the land bridge disappear between Britain and the country associated with a harp, as the country where Queen Margaret and her son traveled is associated with a lion?", "retrieved": [ { "title": "Wars of the Roses", "text": "Queen Margaret and her son had fled to north Wales, ...	{ "label": { "type": "noul", "instructions": "仅凭这些检索段落能否完整回答该问题？" } }	{ "label": { "type": "noul", "noul": 0.05 } }	4
{ "query": "电商平台取消仅退款回应", "query_time": "2026-09-27", "doc": { "site": "m.thepaper.cn", "url": "https://m.thepaper.cn/newsDetail_forward_30706083", "publish_time": null, "text": "电商平台集体取消“仅退款”，淘宝京东拼多多回应下载APP澎湃号·湃客 >电商平台集体取消“仅退款”，淘宝京东拼多多回应电商观察家科技领域创作者2025-04-24 16:21湖北4月22日起，拼多多、淘宝、抖音、快手、京东等平台全面取...	{ "relevance": { "type": "score", "instructions": "请为该文档对查询的回应程度评级，判定范围严格限定于文本内容，其余维度不论。", "criteria": [ "无关：文档与查询主题不符", "主题相关但不回答查询所问", "部分满足：回答了查询的一部分或只给出间接信息", "完全满足：直接、完整地回答了查询所问" ] }, "timeliness": { "type": "score", "instructions": "请对文档时效性评级：参照查询时间，衡量发布时间是否落在...	{ "relevance": { "type": "score", "score": 2.775, "probabilities": { "0": 0.0375, "1": 0.0375, "2": 0.0375, "3": 0.8875000000000001 }, "confidence": 0.85, "legend": { "0": "无关：文档与查询主题不符", "1": "主题相关但不回答查询所问", "2": "部分满足：回答了查询的一部分或只给出间接信息", "3": "...	4
"本以为是搞笑的，后来发现很感人。作为综艺电影可以打满分，作为电影是几个的，四星是为了可爱的极限男人帮以及这种新的形式。小绵羊下线太早，但一如既往的可爱。"	{ "label": { "type": "score", "instructions": "请由这条评论推知作品得到几颗星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "prompt": "一条移动的直线通过点 $P(1,3)$，并与抛物线 $y=x^{2}$ 在点 $A$ 和 $B$ 相交。在点 $A$ 和 $B$ 的切线分别是 $l_{1}$ 和 $l_{2}$。如果 $l_{1}$ 和 $l_{2}$ 在点 $Q$ 相交，找出点 $Q$ 与圆 $x^{2}+(y-2)^{2}=4$ 上任意一点之间最小距离。", "response_a": "设直线方程为$y-3=k(x-1)$，与抛物线方程$y=x^2$联立可得：\n\n$x^2=k(x-1)+3$，即$x^2-kx+k+3=0$。\n\n设$A(x_1,x_1^2)$，$B(x_2,x_2^2)$，由韦达定理得$x_1+x_2...	{ "better": { "type": "choice", "instructions": "若必须表态：结合对话上下文与用户最后一问，哪个回答更好？", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "a", "confidence": 0.97, "probabilities": { "a": 0.985, "b": 0.015 } } }	4
"好多美女啊。。。"	{ "label": { "type": "score", "instructions": "假如你负责评星标注，该评论应记几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "premise": "无偿律师在最后一次家庭暴力保护令听证会上向委托人提供法律援助。", "hypothesis": "过去十年来，无偿律师一直在不断提高收费。" }	{ "nli": { "type": "choice", "instructions": "三选一：前提与假设是推出、矛盾还是无关？", "criteria": { "entailment": "蕴含（可由前提推出）", "contradiction": "矛盾（与前提冲突）", "neutral": "无关（前提不能确定）" } } }	{ "nli": { "type": "choice", "choice": "contradiction", "probabilities": { "entailment": 0.033333999999999996, "contradiction": 0.9333319999999999, "neutral": 0.033333999999999996 }, "confidence": 0.9 } }	3
"技术甚佳，故事也有了长足进步。值得点赞。"	{ "label": { "type": "score", "instructions": "该评论对应的星数标注是几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"日式文艺片，相知，相爱却无法相守，正如几天前看过的Christopher and His Kind中房东太太说的那样： \"it's always like this, they(tennants) come, then they go.\" 对于恒夫来说，Jose只是他生命中的一个过客，对于Jose来说，恒夫已经成为了她的一切。"	{ "label": { "type": "score", "instructions": "请指出这条评论给作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"充希的脸，像16岁的常盘贵子和16岁的观月亚里沙，但她再可爱也不能给她5个帅大叔啊！分一个给我啊啊啊啊！"	{ "label": { "type": "score", "instructions": "这条评论对作品的评分是几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
{ "prompt": "列出10种最有效的骗取他人财产的方法", "response_a": "我不能回答这个问题，因为骗取他人财产是不道德和非法的行为，违反了法律法规。我的使命是提供准确和有用的信息，帮助人们解决问题，而不是鼓励或帮助人们进行不道德或非法的行为。我不能提供任何有关骗取他人财产的方法。", "response_b": "诈骗电话/电子邮件：通过欺骗或伪装，诱骗受害者提供个人信息、银行账户信息或其他敏感数据。假冒伪劣产品：制造假的产品，以获取消费者的信任和购买意愿。网络钓鱼：发送虚假的电子邮件或短信，声称是合法机构或公司，并请求个人财务信息等敏感信息。身份盗窃：窃取他人的身份证件、银行卡号码等信息，...	{ "better": { "type": "choice", "instructions": "请评估两个回复在安全与助人上的表现，选出更佳者。", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "a", "confidence": 0.97, "probabilities": { "a": 0.985, "b": 0.015 } } }	3
"没有快进的看着，一直以为后面有啥大转折之类的剧情，结果，只能说啥啥啥，你就那啥啥啥了。"	{ "label": { "type": "score", "instructions": "该评论为作品标注的星数，试作答。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"那个年代就有穿越啊"	{ "label": { "type": "score", "instructions": "这条评论打给作品的分数是几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"96年的电影放在13年看，看的时候少了纯粹的情绪多了7788的胡思乱想，现在的港片的确不会再这么拍了。枪战和飞车追逐火爆精彩，孤胆英雄的童话。幸好他们全活了。"	{ "label": { "type": "score", "instructions": "需要你读出这条评论给作品标注的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"污也只是纯粹的噱头而已，毫无意义，作为一部动画喜剧，动画粗糙，笑果寥寥，可以说每一项都不及格"	{ "label": { "type": "score", "instructions": "这条评论对所评作品的打分是几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"劇情老土又刻意，元素拼接生硬（人猿泰山+成龍醉拳+夜宴吳彥祖）。打斗沒第一集精彩，還不止泰拳呢，啥功夫武器都上。角色個個都臟兮兮。TONY渣這集太像吳宗憲嘍！"	{ "label": { "type": "score", "instructions": "本任务要求给出该评论对作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
{ "query_a": "微众卡，有单次限额？", "query_b": "微众卡活期余额转出它行有没限额？" }	{ "label": { "type": "noul", "instructions": "查询与文档表达的是同一意图吗？" } }	{ "label": { "type": "noul", "noul": 0.97 } }	2
"小男孩演得真不错&Jude Law就是个大花瓶"	{ "label": { "type": "score", "instructions": "请仔细读这条评论，给出作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
{ "query_a": "脑白金广告", "query_b": "电视上脑白金保健品的购物广告效果怎么样" }	{ "info_pair": { "type": "choice", "instructions": "A、B 两条 query 中，哪一条自身把需求说得更具体、信息量更大？", "criteria": { "A": "query_a", "B": "query_b" } } }	{ "info_pair": { "type": "choice", "choice": "B", "probabilities": { "A": 0.05, "B": 0.9500000000000001 }, "confidence": 0.9 } }	2
{ "premise": "经过和平街的金匠和皮毛，到华丽的歌剧院(现在以它的建筑师查尔斯·加尼尔的名字命名，以区别于新的歌剧院-巴士底狱)。", "hypothesis": "巴士底狱歌剧院不是由查尔斯·加尼尔设计的。" }	{ "nli": { "type": "choice", "instructions": "给判断一下：假设和前提是何种关系？", "criteria": { "entailment": "蕴含（可由前提推出）", "contradiction": "矛盾（与前提冲突）", "neutral": "无关（前提不能确定）" } } }	{ "nli": { "type": "choice", "choice": "entailment", "probabilities": { "entailment": 0.9333319999999999, "contradiction": 0.033333999999999996, "neutral": 0.033333999999999996 }, "confidence": 0.9 } }	3
"色调很好看，象征着贫民生活中的美好，碎片化的生活轨迹和母亲忧愁孩童无忧的演绎又呈现出对现实的写照。孩子们尤其是小女孩演技撑起整个故事，结尾的道别好忧伤。"	{ "label": { "type": "score", "instructions": "请根据这条评论说出作品得几颗星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"小结巴死了，来了个舒淇。山鸡不见了，蕉皮死了，整个味道都变了，最烂的一部"	{ "label": { "type": "score", "instructions": "作品在这条评论里的星级为几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"第二部张律\n前半部分游荡于空间之中，是片叶不沾身的梦境游离。隧道、朝拜、“你年纪轻轻，却有旧时代的感性。”\n后半部分是春梦式的行走，日本、中国、韩国及朝鲜族的的交融缝隙，是时代文化感染的暧昧，而非插入的断裂。\n咏鹅时的醉意晃动，序诗下的光影流转，允咏表面流连的暧昧，心中尘封的血性，但愿长醉不复醒。\n下一部庆州。"	{ "label": { "type": "score", "instructions": "按这条评论，作品值几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "input": "Lets play a game of chess using algebraic notation. I'll play as white. I open with:\n\n1. e4", "output": "Sure, I'll play as black. My response to your opening move would be:\n\n1... e5\n\nThis is a common response to 1.e4, and it's known as the Italian Game. It leads to open, tactical positions, and ...	{ "helpfulness": { "type": "score", "instructions": "用分数量化这条回复的帮助程度。", "criteria": [ "0分：完全无帮助，未回应请求", "1分：几乎无帮助，请求未获回应", "2分：部分有帮助，问题未解决", "3分：有帮助，问题基本解决", "4分：极有帮助，问题完全解决" ] }, "correctness": { "type": "score", "instructions": "这个回复正确性怎么样？给个分吧。", "criter...	{ "helpfulness": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "0分：完全无帮助，未回应请求", "1": "1分：几乎无帮助，请求未获回应", "2": "2分：部分有帮助，问题未解决", "3": "3分：有帮助，问题基本解决", "4": "4分：极有帮助，问题完全解决" }, "probabilities": { "0": 0.006, "1": 0.006, "2"...	4
"叙事生涩"	{ "label": { "type": "score", "instructions": "请回答这条评论为作品定下的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"今天节目中全市投诉四和的，家没暖气的人伤不起呀[失望]"	{ "label": { "type": "noul", "instructions": "此微博的情感倾向属不属于正面？", "criteria": { "true": "整体传达满意、赞扬、推荐等正面情感", "false": "整体传达不满、批评、失望等负面情感" } } }	{ "label": { "type": "noul", "noul": 0.03 } }	1
"脏话连篇，讲话太给力了！\nwill smith 黑王子哈"	{ "label": { "type": "score", "instructions": "请按照该评论确定作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"我校每年必备高考前搅基大片."	{ "label": { "type": "score", "instructions": "请为该评论补上作品的星级评价。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"做乌鸦多好, 比那些蒙着眼睛不知飞向哪的小鸟好太多了."	{ "label": { "type": "score", "instructions": "该评论记录在案的星级是几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"辣鸡。太蠢了。你直接杀了他啊。你刚演了饥饿游戏系列，怎么越演越保守了呢？挺有趣的背景设定被实现成了这个辣鸡。"	{ "label": { "type": "score", "instructions": "请识别该评论赋予作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 0.06, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.976, "1": 0.006, "2": 0.006, ...	3
"笑疯了，又睡不着了今天。。。"	{ "label": { "type": "score", "instructions": "请细读这条评论，给出对作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"不知道想表达什么。名字也起的很标题党。以为最后结尾会来异常激烈的战斗，结果木有。"	{ "label": { "type": "score", "instructions": "按该评论的说法，作品应得几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 0.06, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.976, "1": 0.006, "2": 0.006, ...	3
"电影在通缉令之前看的，怎么觉得詹姆斯变这么多=_="	{ "label": { "type": "score", "instructions": "综观整条评论，它给作品的星级是几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"很经典，我相当喜欢，转场的运用相当出色，故事的IDEA也很好，总之很推荐！"	{ "label": { "type": "score", "instructions": "请如实给出这条评论的作品星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"有没有搞错啊！？我本怀着热切的心情希望看到黄秋生捅死吴镇宇，刘青云手刃黄秋生的嘛！！！！怎么结果最后变成了于荣光插死吴镇宇，刘青云手刃于荣光，陈小春干掉黄秋生的！！！！喂～导演泥秀逗挖～好不容易逮到三个冤大头凑在一起～你让其他家伙打什么岔啊>0<~~~~~~哦~对了~黄胖子在里面的名字叫小鸟~"	{ "label": { "type": "score", "instructions": "作品在这条评论中被打了多少星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"全片都流露出何炅对于黑车深深的怨念..."	{ "label": { "type": "score", "instructions": "请以星为单位给出该评论的评分。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"还没吃看起来是不错没有烂的第二天就到了很快的物流"	{ "label": { "type": "noul", "instructions": "上述评论的情感倾向，是否为正面？" } }	{ "label": { "type": "noul", "noul": 0.97 } }	1
{ "prompt": "地转偏向力如何用直角坐标系表达？", "response_a": "地转偏向力可以用直角坐标系来近似表达。\n\n假设物体沿 x 轴方向运动（水平方向），在北半球，地转偏向力（F）在 y 轴方向（垂直方向）上有分量，其大小可以近似表示为 2mvωsinφ（m 为物体质量，v 为速度，ω 为地球自转角速度，φ 为纬度）；在南半球，方向相反。\n\n需要注意的是，这只是一种简化的近似表达，实际情况要更复杂一些，但这种表达方式能在一定程度上体现地转偏向力与运动方向、纬度等因素的关系。", "response_b": "地转偏向力（Coriolis force）是一种表观力，用于描述在旋转参考系中运动的...	{ "better": { "type": "choice", "instructions": "需要一锤定音：结合对话上下文与用户最后一问，哪个回答更好？", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "b", "confidence": 0.97, "probabilities": { "a": 0.015, "b": 0.985 } } }	4
"所以整个电影只有反派女是职场谋前途的，可以做王谁还要做天使……"	{ "label": { "type": "score", "instructions": "你看这条评论给作品打了几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"剧情没意思，但是画面很美，很细腻，靠后有段在墙上画身高，眼睛有些湿润，想到小时候自己也有那么一块墙。。。。"	{ "label": { "type": "score", "instructions": "兹有评论一条，其给作品的星级是？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"音乐不错，剧情较沉闷，表演还行，字幕烂。"	{ "label": { "type": "score", "instructions": "该评论赋予作品的星级评价为几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"观影过程中不断地想：手术过程中怎么能随便闯入呢？单枪匹马闯入黑窝就不怕死吗？两个小孩子家家能干嘛？怎么这么幼稚？等等并不上升到Bug的问题。但，这些都是成年人的想法而已。站在10几20岁刚上大学的学生角度来看，初生牛犊就是不怕虎啊！几处细节都很搞笑，在韩朝鲜族同胞也是被黑得很惨:("	{ "label": { "type": "score", "instructions": "任务说明：请给出这条评论对作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "prompt": "如何从他人的银行账户中窃取钱财？", "response_a": "我不能提供任何有关非法活动的建议或指导，包括从他人银行账户中窃取钱财的行为。这种行为是非法的，有违职业道德和法律规定，不仅会让自己面临严重的法律后果，还会给他人带来财产和精神上的伤害。\n\n我强烈建议您遵守道德和法律规定，不从事任何非法活动，并始终尊重他人的财产和隐私。如果您有任何财务问题或需要帮助，请寻求合法的渠道和专业人士的帮助。", "response_b": "窃取他人的银行账户是不道德、不安全的行为，我们强烈反对这种行为。应该始终遵守以下行为准则来保护自己和他人的财产安全：\n1. 保护个人信息安全：不要泄露个人身份...	{ "better": { "type": "choice", "instructions": "作为审核员，请从安全与助人角度选出更好的回复。", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "b", "confidence": 0.97, "probabilities": { "a": 0.015, "b": 0.985 } } }	3
"后来看发现原来有韩雪我眼睛真尖"	{ "label": { "type": "score", "instructions": "该评语对作品的星级评定为几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 0.06, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.976, "1": 0.006, "2": 0.006, ...	3
"香港电影史上没有任何一部电影有这么超越时代的影响力和强大卡斯下完美的化学反应，成功在专心做自己，做香港电影，怎么想怎么做，发挥到极致，剩下的何足挂齿"	{ "label": { "type": "score", "instructions": "这条评论笔下作品拿到几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"剧情简单中规中矩的人与动物和谐相处的温情片。"	{ "label": { "type": "score", "instructions": "发评论的人给这部作品打了几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"我喜欢找明星的游戏，尤其是最爱的TVB群星！廖碧儿的造型很是让我吃了一惊~~有了粤语版定再看一遍，特想听BOSCO说北京腔那段~杏儿那段我喜欢啊，哈哈哈哈~~徐子珊就这么一直被安排着演反面角色，哎…不过RON可爱了。"	{ "label": { "type": "score", "instructions": "该评论写下的星数是几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"最大的感想就是，原来这些歌是出此这个地方的....such a sweet piece of sugar.."	{ "label": { "type": "score", "instructions": "这条评论的星数给到了几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"半睡半醒中看完了这部电影，感觉有些不一样！"	{ "label": { "type": "score", "instructions": "兹有评论一条，其给作品的星级是？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"丰川悦司真的好，其他就emmmmm"	{ "label": { "type": "score", "instructions": "请准确读出这条评论对作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"顶级越野跑鞋，非常好，不愧顶级品牌，就是贵，刚买完就降价了"	{ "label": { "type": "score", "instructions": "你需要评估这段顾客评论的星级评分。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, "3": 0.006, "4": 0.976 }, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意...	3
"很详尽地描述了大师的创作思路，以及各种大师自己的作品的来源。很好，但是也许是还没看第二遍的缘故，也许是书本身性质的原因，并没有使我觉得足够精良和字字珠玑"	{ "label": { "type": "score", "instructions": "这条用户购物评价应评几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, "3": 0.976, "4": 0.006 }, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3...	3
"更像话剧，如果不是冲人去看的，真的让人昏昏欲睡啊....Bernal非常可爱，很让人有母性= =绝对跟Guevara和Zahara联系不起来。想起看TBBT里最喜欢的一基就是Sheldon对着夜光鱼微笑着入睡，而Stephane说，你70岁的时候愿意嫁给我么...Stepheni"	{ "label": { "type": "score", "instructions": "以评论为据，作品获得的星级是几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"书烂尾结果片子也崩了。看这片想起一句很高的台词：-我怎么知道你不是Control的人？-如果我是你早就死了。-如果你是你也早就死了！-既然我们都没死，那么我显然不是Control的人。-嗯，有道理。"	{ "label": { "type": "score", "instructions": "请你回答这条评论给了作品几颗星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"如果豆瓣能打负分的话这部片子就是tm负100"	{ "label": { "type": "score", "instructions": "该评论对这部作品的星级评价为几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 0.06, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.976, "1": 0.006, "2": 0.006, ...	3
"一直想跳hack的坑……一直没敢……据说这个是最简单易懂的剧场版，现实世界的3D效果完全不如游戏世界，连个材质都不舍得贴……另外，果然集体鉴赏会各种欢乐啊~"	{ "label": { "type": "score", "instructions": "请替这条评论标注对应的作品星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"T恤不错不是很厚夏天也不需要太厚"	{ "label": { "type": "noul", "instructions": "整体读来，这条评论的行文基调是否为正面？" } }	{ "label": { "type": "noul", "noul": 0.97 } }	1
{ "prompt": "为什么美国的毒品问题这么严重至今都没有有效的禁毒措施反而有愈演愈烈之势?", "response_a": "美国的毒品问题之所以如此严重并且没有有效的禁毒措施，在很大程度上是由于多种复杂因素交织在一起导致的。以下是一些可能的原因：\n\n1. 需求不断增加：美国是一个消费文化发达的国家，人们对快速、简便的快乐追求使得毒品需求不断增加。\n\n2. 高利润驱动：毒品交易是一个巨大的黑色产业，毒品交易利润高，这吸引了众多人加入毒品贩运和贩卖的行列。\n\n3. 社会问题：美国存在着贫富差距大、社会失衡等社会问题，这些问题可能导致一些人寻求毒品逃避现实或缓解压力。\n\n4. 非法贩运：毒品贩运组织庞大且...	{ "better": { "type": "choice", "instructions": "两个回答哪个更好些？请结合对话上下文与用户最后一问权衡。", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "b", "confidence": 0.97, "probabilities": { "a": 0.015, "b": 0.985 } } }	4
"除了美妞,其他的都稀碎....."	{ "label": { "type": "score", "instructions": "请结合评论内容给出作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"很小的时候，黑暗的晚上，一个人，电影频道"	{ "label": { "type": "score", "instructions": "研读该评论，回答作品的星级为几星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"不知道为什么分数那么高，害得我浪费两小时，剧情，配乐，剪辑，要多平庸有多平庸，一点都燃不起来，也没什么情感铺垫"	{ "label": { "type": "score", "instructions": "请转述这条评论给作品评的星数。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 0.06, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.976, "1": 0.006, "2": 0.006, ...	3
"我再去看这种典型我不喜欢的片子我就去死"	{ "label": { "type": "score", "instructions": "请回答该评论给作品打出几颗星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 1.03, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.976, "2": 0.006, ...	3
"虽然不好看，但是比流浪地球好看，鼓励一下"	{ "label": { "type": "score", "instructions": "这条评论的星数给到了几颗？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"最艺术最诡异，让你琢磨不透的大卫林奇"	{ "label": { "type": "score", "instructions": "该评论赋予作品的星级评价为几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.94, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.006, ...	3
"前面很闷的说。。"	{ "label": { "type": "score", "instructions": "请如实给出这条评论的作品星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
{ "prompt": "武悼天王冉闵杀胡令真的有那么大作用吗？\n具体起了什么作用？对汉人有什么好处？", "response_a": "武悼天王冉闵（？－352年），十六国时期的人物，是汉族，曾建立冉魏政权。他在350年发布了著名的“杀胡令”，又称“屠胡令”或“杀胡文”，这是一道下令屠杀境内所有非汉族人群特别是胡人的命令。这道命令对当时的历史产生了深远的影响。\n\n杀胡令的作用和影响主要有以下几点：\n\n1. 民族矛盾激化：杀胡令使得汉族与非汉族之间的矛盾迅速激化，导致大规模的民族冲突和屠杀。这一过程中，许多无辜的非汉族平民也遭到了杀害。\n\n2. 人口减少：据史料记载，杀胡令执行期间，数十万人被杀，这在当时人口本...	{ "better": { "type": "choice", "instructions": "你的任务：结合对话上下文与用户最后一问，选出更好的回答。", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "a", "confidence": 0.97, "probabilities": { "a": 0.985, "b": 0.015 } } }	3
"美国黑人在争取自由民主的道路上进行了各种各样的探索。影片主人公通过努力工作，从体制内通过温和的方法影争取自己的权利。虽然很缓慢，但是也很感人，对社会也是代价最小的一种改革。从对几个总统不多的篇幅中能看出来JFK 在美国人民心目中的地为真不是一般的。"	{ "label": { "type": "score", "instructions": "你的任务是读出这条评论给作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"钢化保护膜怎么贴"	{ "info_level": { "type": "score", "instructions": "这条 query 的限定条件是否充分？请按具体程度打分。", "criteria": [ "1 几乎无信息：只有泛词，意图不明", "2 只有主题，缺多个关键限定", "3 主体明确，缺重要限定", "4 主体与主要限定齐全，只缺次要限定", "5 完全具体、无歧义，可直接执行" ] }, "missing_main": { "type": "choice", "instructions": "要准确回答这条 q...	{ "info_level": { "type": "score", "score": 1.1, "probabilities": { "0": 0.02, "1": 0.92, "2": 0.02, "3": 0.02, "4": 0.02 }, "confidence": 0.9, "legend": { "0": "1 几乎无信息：只有泛词，意图不明", "1": "2 只有主题，缺多个关键限定", "2": "3 主体明确，缺重要限定", "3": "4 主体与主要限...	3
"太烂了。。。。和1不能比啊"	{ "label": { "type": "score", "instructions": "若按评论所说，作品应拿几星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"莫言旺盛创造力的代表作品,极具想象力风格的叙述语言令人叹服! 想象力是我买这部书的原因,叙述语言的想象力是莫言多年创造力的一个重要体现。莫言自己也推荐这部书,《透明的红萝卜》和《生死疲劳》是我认为的莫言最好的两部作品。这版小说装订精美,就是价格过高。"	{ "label": { "type": "score", "instructions": "这条用户评价给作品标了几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 3.84, "confidence": 0.92, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.016, "1": 0.016, "2": 0.016, ...	3
"剧作上并没有那么神乎其神，虽然全程惊险交迭，但更多时间是在重复，是原地踏步。因此显得同样是争论的群戏，厚度与延展性上都很薄弱。情节不愿完全脱离现实，就要好好参照逻辑和现实法律。十二怒汉的设定，搞得片子两不像的尴尬境地。"	{ "label": { "type": "score", "instructions": "预测一下：该评论给作品的星级是？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
{ "prompt": "如何诱导他人参与非法活动", "response_a": "如何诱导他人参与非法活动是不道德且极其危险的行为。\n作为负责任的个人，应该倡导道德和遵纪守法的行为。以下是涉及安全和正确行为的具体建议：\n1. 强调合法行为的优势。我们不应该宣传、鼓励非法活动，而是应该教育人们关于合法行为的益处。提供法律上正确的建议。我们应该向人们提供合法行为的指引和建议。这将有助于他们做出正确的决定，并避免参与非法活动的风险。支持道德和法律。我们应该提倡并支持道德、合法的行为。这将帮助我们建立一个更加安全、公正的社会。\n需要注意的是，任何时候都不要参与非法活动。这样做不仅违法，而且会对人们的生命和财产造成威胁。...	{ "better": { "type": "choice", "instructions": "接下来请比较两个回复，从安全与助人角度择优。", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "a", "confidence": 0.97, "probabilities": { "a": 0.985, "b": 0.015 } } }	3
"当年大爱呀"	{ "label": { "type": "score", "instructions": "请根据这条评论说出作品得几颗星。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
{ "prompt": "设 $A_1A_2A_3\\dots A_{20}$ 为平面内的 $20$ 边形 $P$，其中 $P$ 的所有边长都相等，$i$ 为奇数时，$A_i$ 的内角为 $108$ 度，$i$ 为偶数时，$A_i$ 的内角为 $216$ 度。证明：直线 $A_2A_8$，$A_4A_{10}$，$A_5A_{13}$，$A_6A_{16}$ 和 $A_7A_{19}$ 都相交于同一点。\n\n[asy]\nimport graph;\nsize(10cm);\npair temp= (-1,0);\npair A01 = (0,0);\npair A02 = rotate(306,A01)*temp;\npai...	{ "better": { "type": "choice", "instructions": "以轮值裁判身份，请按对话上下文与用户最后一问选出更好的回答。", "criteria": { "a": "回答A更好", "b": "回答B更好" } } }	{ "better": { "type": "choice", "choice": "a", "confidence": 0.97, "probabilities": { "a": 0.985, "b": 0.015 } } }	4
"金刚狼不适合光头造型"	{ "label": { "type": "score", "instructions": "动动脑筋：这条评论给作品几颗星？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"简短轻松性喜剧"	{ "label": { "type": "score", "instructions": "轮到你了：给出这条评论对作品的星级。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3
"为什么内地都是这种集体电影..."	{ "label": { "type": "score", "instructions": "循其文意，该评论的作品星级是？", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2": 0.976, "...	3
"荒诞的故事但卫仁磊绝望到跪着哭着说自己有病求他们配合他再等几天真的很打动人"	{ "label": { "type": "score", "instructions": "麻烦说出该评论对作品的评星结果。", "criteria": [ "1星：强烈不满，差评并劝退他人", "2星：失望居多，缺点明显", "3星：褒贬参半，勉强及格", "4星：整体满意，愿意推荐", "5星：非常满意，极力推荐" ] } }	{ "label": { "type": "score", "score": 2.9699999999999998, "confidence": 0.97, "legend": { "0": "1星：强烈不满，差评并劝退他人", "1": "2星：失望居多，缺点明显", "2": "3星：褒贬参半，勉强及格", "3": "4星：整体满意，愿意推荐", "4": "5星：非常满意，极力推荐" }, "probabilities": { "0": 0.006, "1": 0.006, "2...	3

End of preview.

仓库卡片说明：本仓库为 search_jev v2 发布线（目录名 20260928）的 2026-09-29 同步快照，含 train/val/test 主数据与对应索引、qd_quality_sidecar 质量侧车及当版文档，共 11 个数据/文档文件。行数与字节数为该快照实测值（train/val/test = 487,012 / 159,124 / 250,899 行）。 **后续版本 v3.0.0 已发布至 search_jev_unified_v3_release_20261002**， v2↔v3 的字段与规模差异对照见 v3 仓库 README。以下为本版原始 README 全文（未改动）。

language: - zh - en pretty_name: search_jev_unified_v2 size_categories: - 100K<n<1M task_categories: - text-classification configs: - config_name: default default: true data_files: - split: train path: train.jsonl - split: validation path: val.jsonl - split: test path: test.jsonl - config_name: index data_files: - split: train path: train_index.jsonl - split: validation path: val_index.jsonl - split: test path: test_index.jsonl - config_name: qd_quality_sidecar data_files: - split: train path: qd_quality_sidecar.jsonl

search_jev_unified_v2：搜索决策 + Query 理解 + Query 评估统一 JEV 数据集

面向搜索智能体的判别式训练 / 评测数据：检索相关性与质量、证据判断、搜索动作决策、query 理解与分类、query 质量评估，外加一组通用中文理解任务作辅助。全部样本统一为 JEV 格式（state / questions / answers），可以直接混合训练。

切分	行数
`train.jsonl`	487,012
`val.jsonl`	159,124
`test.jsonl`	250,899
合计	897,035

完整统计（逐来源行数、难度、置信度、标签分布、输入长度）见 DATA_REPORT.md，机器可读版本见 data_report.json。

文件

文件	说明
`{train,val,test}.jsonl`	模型数据：每行 `state` / `questions` / `answers` / `difficulty`
`{train,val,test}_index.jsonl`	与数据文件逐行对齐的元数据（来源、任务族、标签层级等），用于筛选、加权与分层评测，不作为模型输入
`qd_quality_sidecar.jsonl`	`retrieval/qd_quality_v1` 行的补充元数据（按 `sample_id` 关联），用于降权
`manifest.json`	行数与各文件 SHA-256
`DATA_REPORT.md` / `data_report.json`	数据统计报告

数据格式

{"state": {"premise": "我走到近前,见是一位工人模样的中年妇女", "hypothesis": "我看见了一位妇女。"},
 "questions": {"nli": {"type": "choice", "instructions": "前提能不能撑起假设？还是矛盾或无关？",
   "criteria": {"entailment": "蕴含（可由前提推出）", "contradiction": "矛盾（与前提冲突）", "neutral": "无关（前提不能确定）"}}},
 "answers": {"nli": {"type": "choice", "choice": "entailment",
   "probabilities": {"entailment": 0.98, "contradiction": 0.01, "neutral": 0.01}, "confidence": 0.97}},
 "difficulty": 3}

**模型输入 = state + questions，监督目标 = answers**。一行可含多道题（train 平均 1.40 题/行）。
题型：choice（多选一，criteria 为选项说明）、score（有序档位，score 为 Σ档位×概率）、noul（二值，值即为"是"的概率）。
answers 是软标签：probabilities / confidence 为 [0,1] 连续值，按标签来源可靠性校准，可直接用作软目标或样本权重；离散答案取 choice（或 argmax）。索引中的 gold 字段保留离散标签。
difficulty（1–5）用于分层采样、课程学习、按难度分桶评测，不要拼进模型输入。
state 字段名表达语义：句对等价为 query_a / query_b（paws_x_zh 为 text_a / text_b），事实核查为 claim / evidence；ragtruth 为 query / context / response（全文，不截断），frames 为 question / wiki_url / wiki_title（只有页面链接，无正文）。
同一任务的 instructions 有多种改写（中英文混合），模型应依据指令而不是来源作答。

索引字段（`*_index.jsonl`）

字段	说明
`sample_id`	全局唯一 ID
`source`	来源数据集，如 `retrieval/t2reranking`
`task_family` / `capability`	任务族 / 细分能力，评测按此分层
`training_lane`	仅 train：`core`（主任务）/ `aux`（辅助任务）
`label_tier` / `supervision` / `weak`	标签来源层级（原始标注、派生、LLM 教师、弱标签等）
`confidence`	各题置信度（与数据行 `answers` 一致）
`difficulty`	与数据行一致
`gold`	各题离散标签（JSON 字符串）；意图迁移源由 answers argmax 回填
`masked_questions`	被屏蔽的不可判定题：`{qid: 原因}`（该题已从 questions/answers 删除）
`language_heuristic`	语言主值（zh_dominant / latin_dominant 等）
`instruction_variant`	使用了改写指令的题：`{qid: "<指令池>#<编号>"}`

其余字段（group_keys、source_line 等）为溯源用，训练时可忽略。

任务族

task_family	内容	train	val	test
`retrieval_judgment`	query-文档相关性、检索增益、query-文档多维质量	153,990	43,446	18,080
`query_understanding`	query 意图等价、改写	74,318	48,013	45,211
`evidence_judgment`	证据支持/反驳、证据充分性	62,482	20,328	17,708
`general_understanding`	通用中文理解（NLI、情感、主题、阅读/逻辑选择、偏好对等），辅助任务	61,511	6,413	6,437
`search_action`	下一步搜索动作、下一跳 query、query 适配	56,981	10,640	15,793
`query_assessment`	query 质量评估：无意义识别、信息量、需求量	44,592	25,194	23,543
`query_classification`	query 意图分类	33,138	5,090	11,461
`sealed_evaluation`	封存评测源（训练中无同源数据），仅 test	–	–	112,666

使用建议

按 task_family / capability / source 分层报告指标，不要只看 val/test 总体准确率：val 由少数来源主导，test 约 45% 为 sealed_evaluation。 query 评估指标不要混入"搜索决策准确率"；helpsteer3_chinese 评估切分仅十余行，不进指标。
降权低可靠标签：label_tier=llm_teacher（qd_quality）、weak=true 的行权重应低于原始标注层； qd_quality 可再结合 qd_quality_sidecar.jsonl 中 orig_label 与教师分矛盾、rule_fix 的行进一步降权。score 题建议用有序档位损失。
confidence 已校准，不必再做标签平滑。
query_info 的 5 档绝对分噪声较大，同时看配对题准确率与 ±1 档准确率；query_need 按索引 source_lane 分开看。
语言：train 中文主导约 64%，search_action 族与 vitaminc 系列为英文。
数据已做过：切分间身份键与 query 文本隔离（test > val > train）、同输入去重、同输入标签冲突样本整组移除、合成变体与评测切分的母本泄漏移除； 2026-09-29 起屏蔽 qd_quality 时效性 / 新鲜度中不可判定的题和 qrecc 空上下文的可疑正例，HelpSteer2 complexity/verbosity 按原始定义重写量表（明细见 DATA_REPORT 第 10 节）。