Datasets:

nowzero
/

ESG-Instruct

sample_id stringlengths 18 18	category stringclasses 7 values	task_name stringclasses 35 values	source_dataset stringclasses 13 values	messages listlengths 3 3
train_30k_00000000	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Read the table-derived disclosure and answer the question.\nReturn JSON with this...
train_30k_00000001	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Decide whether the sentence describes the ...
train_30k_00000002	esg_impact_analysis	ml_esg2_jp_to_en__mlesg2_japanese_sentiment_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG sentence sentiment classification assistant. Use only the provided English sentence translated from Japanese. Return valid JSON." }, { "role": "user", "content": "Task: Read the translated sentence and return its ESG sentiment class.\nChoose one of:...
train_30k_00000003	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Identify the action status of the given ES...
train_30k_00000004	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Assign an ESG relevance label to the sentence.\nValid labels: ESG-related, ...
train_30k_00000005	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Assess whether the sentence is materially related to ESG.\nValid labels: ES...
train_30k_00000006	esg_qa	sustainableqa__sustainableqa_non_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer descriptively. Return valid JSON." }, { "role": "user", "content": "Task: Answer without adding outside information beyond the given passage.\nReturn JSON...
train_30k_00000007	esg_impact_analysis	ml_esg3_kr_to_en__mlesg3_impact_duration_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact duration classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Classify the expected ESG impact duration.\nValid labels: Less than 2 years, 2 to 5 years, More than 5 years, Ca...
train_30k_00000008	esg_commitment_analysis	esg_bert_dataset__esg_action_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Identify if the sentence describes ESG implementation, management, monitoring, commitment, or intervention.\nValid labels: yes,...
train_30k_00000009	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Recommend an appropriate gift for a 5 year old girl.\n\nInput:\nNo input" }, { "role": "assistant", "content": "A good gift for a 5 year old girl could be a fun new toy such as a Barbie Dol...
train_30k_00000010	esg_topic_identification	esg_bert_dataset__governance_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Is this text about the Governance dimension of ESG?\nAnswer with yes or no in JSON.\nReturn JSON with this schema: {\"label\": ...
train_30k_00000011	esg_impact_analysis	ml_esg2_jp_to_en__mlesg2_japanese_sentiment_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG sentence sentiment classification assistant. Use only the provided English sentence translated from Japanese. Return valid JSON." }, { "role": "user", "content": "Task: Perform ESG sentence sentiment classification.\nUse exactly one label from Posit...
train_30k_00000012	esg_impact_analysis	dynamic_esg__impact_duration_classification	DynamicESG	[ { "role": "system", "content": "You are an ESG news analysis assistant. Use only the provided English headline and article text. Return valid JSON." }, { "role": "user", "content": "Task: Read the article and infer the likely ESG impact duration.\nUse only these duration labels: <2, 2~5, >5, Not...
train_30k_00000013	esg_impact_analysis	ml_esg2_jp_to_en__mlesg2_japanese_sentiment_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG sentence sentiment classification assistant. Use only the provided English sentence translated from Japanese. Return valid JSON." }, { "role": "user", "content": "Task: Decide the correct sentiment category for the translated ESG sentence.\nReturn J...
train_30k_00000014	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Read the sustainability disclosure and return the answer.\nReturn JSON with this schema: {\"an...
train_30k_00000015	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Check whether the sentence should be kept for ESG analysis.\nLabels: ESG-re...
train_30k_00000016	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Use the corporate report table excerpt to answer accurately.\nReturn JSON with th...
train_30k_00000017	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Return the table-supported answer to the question.\nReturn JSON with this schema:...
train_30k_00000018	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Evaluate whether the sentence is useful for ESG classification.\nAllowed la...
train_30k_00000019	esg_topic_identification	esg_bert_dataset__governance_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Evaluate whether the text is about governance ESG matters.\nAllowed labels: yes, no.\nReturn JSON with this schema: {\"label\":...
train_30k_00000020	esg_topic_identification	dynamic_esg__esg_category_classification	DynamicESG	[ { "role": "system", "content": "You are an ESG news analysis assistant. Use only the provided English headline and article text. Return valid JSON." }, { "role": "user", "content": "Task: Identify all ESG issue category names discussed in the article.\nValid category names: Related to Environmen...
train_30k_00000021	esg_impact_analysis	ml_esg3_fr_to_en__mlesg3_impact_duration_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact duration classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Classify the ESG impact as short, medium, long, or unclear in duration.\nAllowed labels: Less than 2 years, 2 to...
train_30k_00000022	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "List the types of fruits typically found in the American diet." }, { "role": "assistant", "content": "Fruits typically found in the American diet include apples, oranges, bananas, strawberr...
train_30k_00000023	esg_impact_analysis	ml_esg2_fr_to_en__mlesg2_impact_type_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG impact classification assistant. Use only the provided English news title and article excerpt. Return valid JSON." }, { "role": "user", "content": "Task: Label the article as either an ESG opportunity or an ESG risk.\nAllowed labels: Opportunity, Ri...
train_30k_00000024	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Identify the action status of the given ES...
train_30k_00000025	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Determine if the sentence discusses ESG stakeholders, impacts, governance, ...
train_30k_00000026	esg_claim_verification	promise_eval_fr_to_en__promise_evidence_quality_classification	PromiseEval	[ { "role": "system", "content": "You are an ESG evidence quality assessment assistant. Use only the provided English ESG report passage. Return valid JSON." }, { "role": "user", "content": "Task: Identify whether the evidence is clear enough, unclear, misleading, or not applicable.\nReturn JSON w...
train_30k_00000027	esg_topic_identification	esg_bert_dataset__social_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Assign a social ESG relevance label to the sentence.\nValid labels: yes, no.\nReturn JSON with this schema: {\"label\": string}...
train_30k_00000028	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Use the corporate report table excerpt to answer accurately.\nReturn JSON with th...
train_30k_00000029	esg_impact_analysis	ml_esg3_kr_to_en__mlesg3_impact_duration_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact duration classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Assign the passage to the correct ESG impact duration class.\nClasses: Less than 2 years, 2 to 5 years, More tha...
train_30k_00000030	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "How many players are there in a typical football match?" }, { "role": "assistant", "content": "A typical football match is played between two teams of 11 players each, making a total of 22 ...
train_30k_00000031	esg_qa	sustainableqa__sustainableqa_non_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer descriptively. Return valid JSON." }, { "role": "user", "content": "Task: Answer the report-based question with a grounded explanation.\nReturn JSON with ...
train_30k_00000032	esg_topic_identification	esg_bert_dataset__environmental_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Judge whether the sentence is environmentally material in an ESG analysis.\nAllowed labels: yes, no.\nReturn JSON with this sch...
train_30k_00000033	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Decide whether this sentence is ESG-related or unrelated background informa...
train_30k_00000034	esg_impact_analysis	ml_esg3_kr_to_en__mlesg3_impact_type_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact type classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the text describes an ESG opportunity, risk, or cannot be distinguished.\nReturn JSON with this sc...
train_30k_00000035	esg_commitment_analysis	promise_eval_fr_to_en__promise_status_classification	PromiseEval	[ { "role": "system", "content": "You are an ESG promise analysis assistant. Use only the provided English ESG report passage. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the ESG report passage contains a promise, target, commitment, or forward-looking pledge.\nVal...
train_30k_00000036	esg_claim_verification	promise_eval_fr_to_en__promise_evidence_status_classification	PromiseEval	[ { "role": "system", "content": "You are an ESG evidence detection assistant. Use only the provided English ESG report passage. Return valid JSON." }, { "role": "user", "content": "Task: Analyze the ESG report text and classify evidence presence.\nAllowed labels: Yes, No.\nReturn JSON with this s...
train_30k_00000037	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the sentence contains ESG disclosure information rather t...
train_30k_00000038	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: For the given ESG aspect, return exactly o...
train_30k_00000039	esg_topic_identification	esg_bert_dataset__social_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Identify if the sentence describes labor, community, customer, product, or human rights ESG content.\nValid labels: yes, no.\nR...
train_30k_00000040	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Answer the report table question with information from the context only.\nReturn ...
train_30k_00000041	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Use the given corporate report excerpt to answer accurately.\nReturn JSON with this schema: {\...
train_30k_00000042	esg_commitment_analysis	promise_eval_jp_to_en__promise_verification_timeline_classification	PromiseEval	[ { "role": "system", "content": "You are an ESG promise timeline classification assistant. Use only the provided English ESG report passage. Return valid JSON." }, { "role": "user", "content": "Task: Perform ESG promise timeline classification.\nUse exactly one label from: Already, Less than 2 ye...
train_30k_00000043	esg_topic_identification	esg_bert_dataset__social_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Does this sentence discuss social topics such as employees, communities, human rights, safety, diversity, customers, or supplie...
train_30k_00000044	esg_qa	sustainableqa__sustainableqa_non_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer descriptively. Return valid JSON." }, { "role": "user", "content": "Task: Provide a context-grounded explanatory answer to the question.\nReturn JSON with...
train_30k_00000045	esg_impact_analysis	ml_esg3_jp_to_en__mlesg3_impact_duration_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact duration classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Perform ESG impact duration classification.\nChoose exactly one label: Less than 2 years, 2 to 5 years, More tha...
train_30k_00000046	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Provide the precise answer supported by the context.\nReturn JSON with this schema: {\"answer\...
train_30k_00000047	esg_impact_analysis	ml_esg3_jp_to_en__mlesg3_impact_type_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact type classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Perform ESG impact type classification.\nUse exactly one label: Opportunity, Risk, or Cannot distinguish.\nReturn JS...
train_30k_00000048	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Inspect the sentence and output the correc...
train_30k_00000049	esg_qa	sustainableqa__sustainableqa_non_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer descriptively. Return valid JSON." }, { "role": "user", "content": "Task: Answer without adding outside information beyond the given passage.\nReturn JSON...
train_30k_00000050	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Identify if the sentence contains environmental, social, governance, or oth...
train_30k_00000051	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Generate a dialogue between two characters, Jack and Susan, in a restaurant." }, { "role": "assistant", "content": "Jack: Hey, this place looks really nice. How long have you been coming he...
train_30k_00000052	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Decide whether the sentence describes the ...
train_30k_00000053	esg_topic_identification	a3cg_greenwash__aspect_extraction	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Extract the ESG aspect terms associated wi...
train_30k_00000054	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Decide whether this corporate disclosure sentence discusses an ESG matter.\...
train_30k_00000055	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Provide the answer supported by the table context only.\nReturn JSON with this sc...
train_30k_00000056	esg_topic_identification	esg_activities__esg_activity_concept_matching	ESG_activities	[ { "role": "system", "content": "You are an ESG taxonomy matching assistant. Use only the provided activity text and ESG key concept. Return valid JSON." }, { "role": "user", "content": "Task: Classify whether the activity maps to the ESG key concept.\nLabels must be selected from: 1, 0.\nReturn ...
train_30k_00000057	esg_impact_analysis	ml_esg2_fr_to_en__mlesg2_impact_type_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG impact classification assistant. Use only the provided English news title and article excerpt. Return valid JSON." }, { "role": "user", "content": "Task: Assign the article to the ESG impact type that best describes its content.\nImpact labels: Oppo...
train_30k_00000058	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Determine the correct answer to the question from the context.\nReturn JSON with this schema: ...
train_30k_00000059	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Locate the requested entity, value, or phrase in the context.\nReturn JSON with this schema: {...
train_30k_00000060	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the sentence contains ESG disclosure information rather t...
train_30k_00000061	esg_impact_analysis	ml_esg2_jp_to_en__mlesg2_japanese_sentiment_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG sentence sentiment classification assistant. Use only the provided English sentence translated from Japanese. Return valid JSON." }, { "role": "user", "content": "Task: Classify the sentence's ESG sentiment polarity.\nAllowed labels: Positive, Negat...
train_30k_00000062	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Judge whether the sentence should be treated as ESG-relevant in a Korean su...
train_30k_00000063	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Generate a sentence using the provided noun and verb in present perfect tense.\n\nInput:\nNoun: students \nVerb: complete" }, { "role": "assistant", "content": "The students have completed ...
train_30k_00000064	esg_impact_analysis	ml_esg3_jp_to_en__mlesg3_impact_type_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact type classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Classify the content as an ESG opportunity, ESG risk, or unclear.\nAllowed labels: Opportunity, Risk, Cannot disting...
train_30k_00000065	esg_topic_identification	ml_esg1_fr_to_en__mlesg1_esg_issue_classification	ML-ESG1	[ { "role": "system", "content": "You are an ESG news classification assistant. Use only the provided English news title and article excerpt. Return valid JSON." }, { "role": "user", "content": "Task: Perform ESG issue classification for this news item.\nUse exactly one category from this list: Ac...
train_30k_00000066	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Identify whether ESG content is present in the translated sentence.\nValid ...
train_30k_00000067	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Construct a metaphor about intelligence" }, { "role": "assistant", "content": "Intelligence is like a garden. With the right care and nurturing, it can bloom into a beautiful and vibrant la...
train_30k_00000068	esg_impact_analysis	dynamic_esg__impact_duration_classification	DynamicESG	[ { "role": "system", "content": "You are an ESG news analysis assistant. Use only the provided English headline and article text. Return valid JSON." }, { "role": "user", "content": "Task: Decide the expected ESG impact duration for the company or issue discussed.\nValid duration labels are: <2, ...
train_30k_00000069	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Compute the average of the following numbers: 6, 15, 28, 42" }, { "role": "assistant", "content": "23.5" } ]
train_30k_00000070	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the sentence contains ESG disclosure information rather t...
train_30k_00000071	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Classify the sentence as relevant to ESG disclosure or not.\nAllowed labels...
train_30k_00000072	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Determine the correct answer from the table summary.\nReturn JSON with this schem...
train_30k_00000073	esg_impact_analysis	ml_esg3_fr_to_en__mlesg3_impact_duration_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact duration classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Classify the time frame associated with the ESG impact.\nUse one of: Less than 2 years, 2 to 5 years, More than ...
train_30k_00000074	esg_qa	sustainableqa__sustainableqa_table_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability and EU Taxonomy table question-answering assistant. Use only the provided table context to answer. Return valid JSON." }, { "role": "user", "content": "Task: Identify the table value or conclusion needed to answer the question.\nReturn JSO...
train_30k_00000075	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Return the action label associated with th...
train_30k_00000076	esg_topic_identification	esg_bert_dataset__social_classification	ESG_BERT_dataset	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided text. Return valid JSON." }, { "role": "user", "content": "Task: Decide whether this sentence belongs in a social ESG classifier.\nUse yes or no.\nReturn JSON with this schema: {\"label\": str...
train_30k_00000077	esg_qa	sustainableqa__sustainableqa_non_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer descriptively. Return valid JSON." }, { "role": "user", "content": "Task: Answer the descriptive question using only the context.\nReturn JSON with this s...
train_30k_00000078	esg_commitment_analysis	promise_eval_jp_to_en__promise_status_classification	PromiseEval	[ { "role": "system", "content": "You are an ESG promise analysis assistant. Use only the provided English ESG report passage. Return valid JSON." }, { "role": "user", "content": "Task: Check whether a promise, goal, or target is present in the ESG passage.\nUse Yes or No.\nReturn JSON with this s...
train_30k_00000079	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Read the sentence and assign the correct a...
train_30k_00000080	esg_impact_analysis	dynamic_esg__impact_type_classification	DynamicESG	[ { "role": "system", "content": "You are an ESG news analysis assistant. Use only the provided English headline and article text. Return valid JSON." }, { "role": "user", "content": "Task: Use the article to determine whether its ESG implication is opportunity, risk, ambiguous, or unrelated.\nLab...
train_30k_00000081	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Check whether the sentence should be kept for ESG analysis.\nLabels: ESG-re...
train_30k_00000082	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Given a list of natural disasters and their description, generate a poem about them.\n\nInput:\nTsunami: A huge wave caused by the displacement of a large body of water\nWildfire: Fire that burns large...
train_30k_00000083	esg_impact_analysis	ml_esg2_jp_to_en__mlesg2_japanese_sentiment_classification	ML-ESG2	[ { "role": "system", "content": "You are an ESG sentence sentiment classification assistant. Use only the provided English sentence translated from Japanese. Return valid JSON." }, { "role": "user", "content": "Task: Classify the ESG sentiment of the translated Japanese sentence.\nValid labels: P...
train_30k_00000084	esg_impact_analysis	ml_esg3_kr_to_en__mlesg3_impact_type_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact type classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Assign the ESG passage to the correct impact type class.\nClasses: Opportunity, Risk, Cannot distinguish.\nReturn JS...
train_30k_00000085	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Change the text to proper nouns.\n\nInput:\nthe first humans" }, { "role": "assistant", "content": "Adam and Eve" } ]
train_30k_00000086	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Classify the sentence according to ESG relevance.\nValid labels: ESG-relate...
train_30k_00000087	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Decide whether this corporate disclosure sentence discusses an ESG matter.\...
train_30k_00000088	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Read the sustainability disclosure and return the answer.\nReturn JSON with this schema: {\"an...
train_30k_00000089	esg_topic_identification	esg_activities__esg_activity_concept_matching	ESG_activities	[ { "role": "system", "content": "You are an ESG taxonomy matching assistant. Use only the provided activity text and ESG key concept. Return valid JSON." }, { "role": "user", "content": "Task: Judge whether the activity satisfies or clearly relates to the ESG key concept.\nLabels: 1 = related, 0 ...
train_30k_00000090	esg_claim_verification	promise_eval_jp_to_en__promise_evidence_status_classification	PromiseEval	[ { "role": "system", "content": "You are an ESG evidence detection assistant. Use only the provided English ESG report passage. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the ESG passage provides evidence for its promise or claim.\nValid labels: Yes, No.\nReturn ...
train_30k_00000091	esg_impact_analysis	ml_esg3_fr_to_en__mlesg3_impact_level_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact severity classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Label the ESG-related article by impact level.\nValid options: Low, Medium, High.\nReturn JSON with this schema:...
train_30k_00000092	general_instruction_chat	alpaca_general_instruction	tatsu-lab/alpaca	[ { "role": "system", "content": "You are a helpful assistant." }, { "role": "user", "content": "Summarize the main points of the given article using 100 words or less.\n\nInput:\nhttps://www.theverge.com/2021/3/25/22350066/apple-app-tracking-transparency-transparency-nutrition-privacy-ads-on-ios"...
train_30k_00000093	esg_relevance_detection	esg_kor__korean_esg_relevance_classification	ESG_Kor	[ { "role": "system", "content": "You are an ESG sentence classification assistant. Use only the provided English sentence translated from a Korean ESG disclosure. Return valid JSON." }, { "role": "user", "content": "Task: Assign an ESG relevance label to the sentence.\nValid labels: ESG-related, ...
train_30k_00000094	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Determine how the sentence frames the spec...
train_30k_00000095	esg_impact_analysis	ml_esg3_kr_to_en__mlesg3_impact_type_classification	ML-ESG3	[ { "role": "system", "content": "You are an ESG impact type classification assistant. Use only the provided English text. Return valid JSON." }, { "role": "user", "content": "Task: Determine whether the ESG content is opportunity-oriented, risk-oriented, or unclear.\nReturn JSON with this schema:...
train_30k_00000096	esg_qa	sustainableqa__sustainableqa_non_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer descriptively. Return valid JSON." }, { "role": "user", "content": "Task: Use the given corporate report excerpt to answer descriptively and accurately.\n...
train_30k_00000097	esg_topic_identification	dynamic_esg__esg_category_classification	DynamicESG	[ { "role": "system", "content": "You are an ESG news analysis assistant. Use only the provided English headline and article text. Return valid JSON." }, { "role": "user", "content": "Task: Infer the DynamicESG category-name labels from the English translated article.\nAllowed category names: Rela...
train_30k_00000098	esg_qa	sustainableqa__sustainableqa_factoid_answering	SustainableQA	[ { "role": "system", "content": "You are a sustainability report question-answering assistant. Use only the provided context passage to answer. Return valid JSON." }, { "role": "user", "content": "Task: Determine the correct answer to the question from the context.\nReturn JSON with this schema: ...
train_30k_00000099	esg_commitment_analysis	a3cg_greenwash__action_classification	a3cg_greenwash	[ { "role": "system", "content": "You are an ESG information extraction assistant. Follow the task instruction exactly. Use only explicit evidence from the input text. Do not infer unstated facts. Return valid JSON." }, { "role": "user", "content": "Task: Determine how the sentence frames the spec...

End of preview. Expand in Data Studio

ESG-Instruct 30K

ESG-Instruct is an English instruction-tuning dataset for environmental, social, and governance (ESG) tasks. The public 30K release combines ESG classification, extraction, verification, and question-answering examples with a general instruction-following component. Every example is represented as a three-message system/user/assistant conversation, and structured tasks request JSON output.

Dataset file

File	Rows	Description
`esg_instruct_30k.jsonl`	30,000	Public instruction-tuning split across 35 detail tasks

The public release contains the 30K training split only. Internal source pools, alternative-size subsets, sampling analyses, and duplicate-check artifacts are not included.

Loading

from datasets import load_dataset

dataset = load_dataset("nowzero/ESG-Instruct", split="train")

messages = dataset[0]["messages"]
prompt_messages = messages[:-1]
target_answer = messages[-1]["content"]

To load a downloaded copy directly:

dataset = load_dataset(
    "json",
    data_files="esg_instruct_30k.jsonl",
    split="train",
)

Main schema

sample_id: unique identifier in the public 30K split
category, task_name: unified category and task identifiers
source_dataset: source provenance
messages: system, user, and assistant messages; the final assistant message is the supervised target

Composition

Category	Rows
ESG relevance detection	5,160
ESG topic identification	5,170
ESG commitment analysis	4,736
ESG impact analysis	5,170
ESG claim verification	1,600
ESG question answering	5,164
General instruction chat	3,000
Total	30,000

The release contains 35 detail tasks across seven instruction categories. The ESG portion includes relevance and topic classification, activity and aspect extraction, impact type/duration/level and sentiment classification, commitment and evidence analysis, and sustainability QA. The general component consists of 3,000 Alpaca instruction-following examples.

Construction

Records were normalized to a common chat format and sampled from training splits using category, task, and label-aware balancing. The final split contains 30,000 unique message records without oversampling. Sources translated into English may retain translation artifacts.

Intended use and limitations

ESG-Instruct supports research on ESG instruction tuning, multi-task learning, structured prediction, and sustainability QA. It combines datasets with different annotation guidelines, label spaces, domains, and source biases, and automatic conversion may introduce errors. Trained models may hallucinate, overfit output formats, or perform poorly on unseen jurisdictions and reporting standards. The dataset is not a source of financial, legal, regulatory, investment, or compliance advice.

Citation

If you use ESG-Instruct, please cite:

Jaeyoung Lee and Misuk Kim. “ESGPT: A Data-Centric ESG Language Model with Unified Instruction and Benchmark Resources.” In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2026.

@inproceedings{lee2026esgpt,
  title     = {{ESGPT}: A Data-Centric {ESG} Language Model with Unified Instruction and Benchmark Resources},
  author    = {Lee, Jaeyoung and Kim, Misuk},
  booktitle = {Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing},
  year      = {2026}
}

Code: https://github.com/nowzer0/ESGPT

License

ESG-Instruct 30K is released under the Creative Commons Attribution-NonCommercial 4.0 International license (CC BY-NC 4.0). See LICENSE for details.

Downloads last month: 34

Models trained or fine-tuned on nowzero/ESG-Instruct

nowzero/ESGPT

Text Generation • Updated 20 days ago • 13