{"id":"https://openalex.org/W4416034701","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.451","title":"Staged Knowledge Distillation Through Least-to-Most Prompting: Optimizing Teacher Guidance via Difficulty-Aware Training","display_name":"Staged Knowledge Distillation Through Least-to-Most Prompting: Optimizing Teacher Guidance via Difficulty-Aware Training","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034701","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.451"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.451","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.451","pdf_url":"https://aclanthology.org/2025.findings-emnlp.451.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-emnlp.451.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5109763087","display_name":"Mengxiang Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I168719708","display_name":"City University of Hong Kong","ror":"https://ror.org/03q8dnn23","country_code":"HK","type":"education","lineage":["https://openalex.org/I168719708"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Mengxiang Zhang","raw_affiliation_strings":["The University of Hong Kong City University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Hong Kong City University of Hong Kong","institution_ids":["https://openalex.org/I168719708","https://openalex.org/I889458895"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5009327148","display_name":"Lingyuan Liu","orcid":"https://orcid.org/0009-0008-9777-4550"},"institutions":[{"id":"https://openalex.org/I168719708","display_name":"City University of Hong Kong","ror":"https://ror.org/03q8dnn23","country_code":"HK","type":"education","lineage":["https://openalex.org/I168719708"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Lingyuan Liu","raw_affiliation_strings":["The University of Hong Kong City University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Hong Kong City University of Hong Kong","institution_ids":["https://openalex.org/I168719708","https://openalex.org/I889458895"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.15740528,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"8489","last_page":"8501"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.557699978351593,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.557699978351593,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12072","display_name":"Machine Learning and Algorithms","score":0.07919999957084656,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11516","display_name":"Visual and Cognitive Learning Processes","score":0.027899999171495438,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.6279000043869019},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.30149999260902405},{"id":"https://openalex.org/keywords/expert-system","display_name":"Expert system","score":0.27570000290870667},{"id":"https://openalex.org/keywords/knowledge-based-systems","display_name":"Knowledge-based systems","score":0.275299996137619},{"id":"https://openalex.org/keywords/troubleshooting","display_name":"Troubleshooting","score":0.26589998602867126},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.265500009059906}],"concepts":[{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.6279000043869019},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5249999761581421},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.325300008058548},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.32420000433921814},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.30149999260902405},{"id":"https://openalex.org/C58328972","wikidata":"https://www.wikidata.org/wiki/Q184609","display_name":"Expert system","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C115925183","wikidata":"https://www.wikidata.org/wiki/Q1412694","display_name":"Knowledge-based systems","level":2,"score":0.275299996137619},{"id":"https://openalex.org/C147494362","wikidata":"https://www.wikidata.org/wiki/Q2078905","display_name":"Troubleshooting","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.265500009059906},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.2639999985694885},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.2637999951839447},{"id":"https://openalex.org/C2777220311","wikidata":"https://www.wikidata.org/wiki/Q6423340","display_name":"Knowledge acquisition","level":2,"score":0.2574999928474426},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.451","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.451","pdf_url":"https://aclanthology.org/2025.findings-emnlp.451.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},{"id":"pmh:oai:pure.atira.dk:publications/fa5b8350-d24f-4996-884b-17642859b63e","is_oa":true,"landing_page_url":"https://hdl.handle.net/2031/fa5b8350-d24f-4996-884b-17642859b63e","pdf_url":null,"source":{"id":"https://openalex.org/S7407055387","display_name":"CityU Scholars","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Zhang, M & Liu, L 2025, Staged Knowledge Distillation Through Least-to-Most Prompting : Optimizing Teacher Guidance via Difficulty-Aware Training. in Findings of the Association for Computational Linguistics : EMNLP 2025. EMNLP - Conference on Empirical Methods in Natural Language Processing, Findings of EMNLP, Association for Computational Linguistics, pp. 8489-8501, 30th Conference on Empirical Methods in Natural Language Processing (EMNLP 2025), Suzhou, China, 4/11/25. https://doi.org/10.18653/v1/2025.findings-emnlp.451","raw_type":"info:eu-repo/semantics/publishedVersion"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.451","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.451","pdf_url":"https://aclanthology.org/2025.findings-emnlp.451.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416034701.pdf","grobid_xml":"https://content.openalex.org/works/W4416034701.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Knowledge":[0,68],"distillation":[1],"(KD)":[2],"enables":[3],"the":[4,47,160],"compression":[5],"of":[6,50,162],"large":[7],"language":[8],"models":[9],"(LLMs)":[10],"by":[11,97,153],"transferring":[12],"knowledge":[13],"from":[14,39,118],"a":[15,20,71,86,90,170],"high-capacity":[16],"teacher":[17,125,167],"model":[18,147],"to":[19,46,120,127,176],"resource-efficient":[21],"student":[22,146],"model,":[23],"maintaining":[24],"competitive":[25],"performance":[26,44,148],"for":[27],"tasks":[28,135],"such":[29],"as":[30],"instruction":[31],"following.However,":[32],"conventional":[33],"white-box":[34,73,141],"KD":[35,74,114,142],"methods":[36],"often":[37],"suffer":[38],"training-inference":[40],"mismatches":[41],"and":[42,54,81,110,130,165,173],"suboptimal":[43],"due":[45],"asymmetric":[48],"nature":[49],"Kullback-Leibler":[51],"divergence":[52],"(KLD)":[53],"reliance":[55],"on":[56,133],"computationally":[57,171],"expensive":[58],"student-generated":[59],"outputs.To":[60],"address":[61],"these":[62],"challenges,":[63],"we":[64],"propose":[65],"Least-to-Most":[66],"Prompting":[67],"Distillation":[69],"(L2M-KD),":[70],"novel":[72],"method":[75],"grounded":[76],"in":[77],"curriculum":[78],"learning":[79],"(CL)":[80],"adaptive":[82,113,166],"loss":[83,115],"design.L2M-KD":[84],"employs":[85],"two-pronged":[87],"approach:":[88],"(1)":[89],"CL":[91],"strategy":[92],"that":[93,116,137],"ranks":[94],"training":[95,164],"samples":[96],"difficulty":[98],"using":[99],"Rouge-L":[100],"scores,":[101],"partitioning":[102],"them":[103],"into":[104],"easy-to-hard":[105],"subsets":[106],"across":[107],"multiple":[108],"stages,":[109],"(2)":[111],"an":[112],"transitions":[117],"KLD":[119],"skew":[121],"KLD,":[122],"dynamically":[123],"adjusting":[124],"guidance":[126],"mitigate":[128],"mode-averaging":[129],"over-smoothing.Extensive":[131],"experiments":[132],"instruction-following":[134],"demonstrate":[136],"L2M-KD":[138],"outperforms":[139],"existing":[140],"methods,":[143],"achieving":[144],"superior":[145],"with":[149],"reduced":[150],"computational":[151],"overhead":[152],"leveraging":[154],"ground-truth":[155],"outputs":[156],"exclusively.Our":[157],"findings":[158],"underscore":[159],"efficacy":[161],"difficultyaware":[163],"guidance,":[168],"offering":[169],"efficient":[172],"robust":[174],"approach":[175],"LLM":[177],"compression.":[178]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-11-08T00:00:00"}