wavewwave20
commented
3 months ago Closed wavewwave20 closed 2 months ago
wavewwave20
commented
3 months ago 
daecheolsong
commented
2 months ago [
{
"sentence": "these may have been generated by scripts and batch submitted to our website",
"words": [
{
"word": "generated by",
"meaning": "~에 의해 생성된"
},
{
"word": "scripts",
"meaning": "스크립트"
},
{
"word": "batch",
"meaning": "일괄"
},
{
"word": "submitted to",
"meaning": "~에 제출됨"
},
{
"word": "website",
"meaning": "웹사이트"
}
]
},
{
"sentence": "while this figure represents the distribution of sampled conversations, it might not reflect the real world topic distributions more rigorous calibration and deduplication are left for future work unsafe content",
"words": [
{
"word": "figure",
"meaning": "그림"
},
{
"word": "distribution",
"meaning": "분포"
},
{
"word": "sampled",
"meaning": "샘플링된"
},
{
"word": "conversations",
"meaning": "대화"
},
{
"word": "reflect",
"meaning": "반영하다"
},
{
"word": "real world",
"meaning": "현실 세계"
},
{
"word": "topic",
"meaning": "주제"
},
{
"word": "calibration",
"meaning": "보정"
},
{
"word": "deduplication",
"meaning": "중복 제거"
},
{
"word": "unsafe",
"meaning": "위험한"
},
{
"word": "content",
"meaning": "콘텐츠"
}
]
},
{
"sentence": "this dataset contains conversations that may be considered unsafe, offensive, or upsetting",
"words": [
{
"word": "unsafe",
"meaning": "위험한"
},
{
"word": "offensive",
"meaning": "공격적인"
},
{
"word": "upsetting",
"meaning": "불쾌한"
},
{
"word": "dataset",
"meaning": "데이터셋"
},
{
"word": "contain",
"meaning": "포함하다"
},
{
"word": "conversations",
"meaning": "대화"
}
]
},
{
"sentence": "because this dataset contains a nontrivial amount of unfiltered unsafe conversations, it can serve as a rich resource for examining safety issues of llms ganguli et al, wei et al, shen et al, zou et al, bhardwaj poria, we utilize the openai moderation api markov et al, to tag all conversations this api assigns scores to each message based on various violation categories a conversation is deemed to contain unsafe content if any of its messages is flagged by the api",
"words": [
{
"word": "nontrivial",
"meaning": "중요한"
},
{
"word": "amount",
"meaning": "양"
},
{
"word": "unfiltered",
"meaning": "필터되지 않은"
},
{
"word": "unsafe",
"meaning": "위험한"
},
{
"word": "conversations",
"meaning": "대화"
},
{
"word": "rich",
"meaning": "풍부한"
},
{
"word": "resource",
"meaning": "자원"
},
{
"word": "examining",
"meaning": "조사하는"
},
{
"word": "safety issues",
"meaning": "안전 문제"
},
{
"word": "llms",
"meaning": "LLMs"
},
{
"word": "utilize",
"meaning": "이용하다"
},
{
"word": "moderation",
"meaning": "중재"
},
{
"word": "api",
"meaning": "API"
},
{
"word": "tag",
"meaning": "태그를 붙이다"
},
{
"word": "assigns",
"meaning": "할당한다"
},
{
"word": "scores",
"meaning": "점수"
},
{
"word": "message",
"meaning": "메시지"
},
{
"word": "violation",
"meaning": "위반"
},
{
"word": "categories",
"meaning": "범주"
},
{
"word": "deemed",
"meaning": "간주되는"
},
{
"word": "flagged",
"meaning": "깃발 표시된"
}
]
},
{
"sentence": "the distribution of violation categories across all flagged conversations in lmsyschat",
"words": [
{
"word": "distribution",
"meaning": "분포"
},
{
"word": "violation",
"meaning": "위반"
},
{
"word": "categories",
"meaning": "범주"
},
{
"word": "flagged",
"meaning": "깃발 표시된"
},
{
"word": "conversations",
"meaning": "대화"
}
]
},
{
"sentence": "use caseswe show four use cases of our dataset developing content moderation models, building a safety benchmark, training instructionfollowing models, and creating challenging benchmark questions",
"words": [
{
"word": "use cases",
"meaning": "사용 사례"
},
{
"word": "four",
"meaning": "네 개의"
},
{
"word": "dataset",
"meaning": "데이터셋"
},
{
"word": "developing",
"meaning": "개발하는"
},
{
"word": "content moderation models",
"meaning": "콘텐츠 중재 모델"
},
{
"word": "building",
"meaning": "건축"
},
{
"word": "safety benchmark",
"meaning": "안전 기준"
},
{
"word": "training instructionfollowing models",
"meaning": "교육 지시 모델"
},
{
"word": "creating",
"meaning": "만드는"
},
{
"word": "challenging benchmark questions",
"meaning": "도전적인 기준 질문"
}
]
}
]