{"id":"https://openalex.org/W4406954331","doi":"https://doi.org/10.48550/arxiv.2501.16673","title":"LLM-AutoDiff: Auto-Differentiate Any LLM Workflow","display_name":"LLM-AutoDiff: Auto-Differentiate Any LLM Workflow","publication_year":2025,"publication_date":"2025-01-28","ids":{"openalex":"https://openalex.org/W4406954331","doi":"https://doi.org/10.48550/arxiv.2501.16673"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2501.16673","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2501.16673","pdf_url":"https://arxiv.org/pdf/2501.16673","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2501.16673","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Yin, Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yin, Li","raw_affiliation_strings":["Atlas"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Atlas","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5048522863","display_name":"Zhangyang Wang","orcid":"https://orcid.org/0000-0002-2050-5693","observed_orcids":["https://orcid.org/0000-0002-2050-5693"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhangyang","raw_affiliation_strings":["Atlas"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Atlas","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":26,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9413999915122986,"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/T10028","display_name":"Topic Modeling","score":0.9413999915122986,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.03700000047683716,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.006500000134110451,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7900000214576721},{"id":"https://openalex.org/keywords/prompt-optimization","display_name":"prompt optimization","score":0.6489999890327454},{"id":"https://openalex.org/keywords/multi-hop-retrieval","display_name":"multi-hop retrieval","score":0.5630000233650208},{"id":"https://openalex.org/keywords/question-answering","display_name":"question answering","score":0.7730000019073486},{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.671999990940094}],"concepts":[{"id":"https://openalex.org/C177212765","wikidata":"https://www.wikidata.org/wiki/Q627335","display_name":"Workflow","level":2,"score":0.8080775737762451},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.38476449251174927},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.32702112197875977},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.14294031262397766}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2501.16673","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2501.16673","pdf_url":"https://arxiv.org/pdf/2501.16673","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2501.16673","doi":"https://doi.org/10.48550/arxiv.2501.16673","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2501.16673","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2501.16673","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2501.16673","pdf_url":"https://arxiv.org/pdf/2501.16673","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4406954331.pdf","grobid_xml":"https://content.openalex.org/works/W4406954331.grobid-xml"},"referenced_works_count":15,"referenced_works":["https://openalex.org/W2952862139","https://openalex.org/W3030163527","https://openalex.org/W4281690148","https://openalex.org/W4304195432","https://openalex.org/W4306808908","https://openalex.org/W4308244910","https://openalex.org/W4309868377","https://openalex.org/W4313304293","https://openalex.org/W4385573003","https://openalex.org/W4386554921","https://openalex.org/W4389520756","https://openalex.org/W4399655920","https://openalex.org/W4404782958","https://openalex.org/W4404792910","https://openalex.org/W4405096496"],"related_works":["https://openalex.org/W4415428098","https://openalex.org/W4389520756","https://openalex.org/W7091625288","https://openalex.org/W7117994624","https://openalex.org/W4416381525","https://openalex.org/W4417262486","https://openalex.org/W4406058569","https://openalex.org/W4409362909","https://openalex.org/W4399794854","https://openalex.org/W4416290380"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4,215],"reshaped":[5],"natural":[6],"language":[7],"processing,":[8],"powering":[9],"applications":[10],"from":[11],"multi-hop":[12,131,167],"retrieval":[13,53],"and":[14,38,54,96,133,170,183,202],"question":[15],"answering":[16],"to":[17,31,76,103,106],"autonomous":[18],"agent":[19],"workflows.":[20],"Yet,":[21],"prompt":[22,113,188],"engineering":[23],"--":[24,35,109],"the":[25,84,135,208],"task":[26],"of":[27],"crafting":[28],"textual":[29,70,90,107,177],"inputs":[30],"effectively":[32],"direct":[33],"LLMs":[34],"remains":[36],"difficult":[37],"labor-intensive,":[39],"particularly":[40],"for":[41,63,200],"complex":[42],"pipelines":[43],"that":[44,68,110,211],"combine":[45],"multiple":[46],"LLM":[47,80,102,204],"calls":[48,129],"with":[49],"functional":[50,122],"operations":[51],"like":[52],"data":[55],"formatting.":[56],"We":[57],"introduce":[58],"LLM-AutoDiff:":[59],"a":[60,93,98,191,196],"novel":[61],"framework":[62],"Automatic":[64],"Prompt":[65],"Engineering":[66],"(APE)":[67],"extends":[69],"gradient-based":[71],"methods":[72],"(such":[73],"as":[74,92],"Text-Grad)":[75],"multi-component,":[77],"potentially":[78],"cyclic":[79],"architectures.":[81],"Implemented":[82],"within":[83],"AdalFlow":[85],"library,":[86],"LLM-AutoDiff":[87,119,173,194],"treats":[88],"each":[89],"input":[91],"trainable":[94],"parameter":[95],"uses":[97],"frozen":[99],"backward":[100],"engine":[101],"generate":[104],"feedback-akin":[105],"gradients":[108],"guide":[111],"iterative":[112],"updates.":[114],"Unlike":[115],"prior":[116],"single-node":[117],"approaches,":[118],"inherently":[120],"accommodates":[121],"nodes,":[123],"preserves":[124],"time-sequential":[125],"behavior":[126],"in":[127,180,218],"repeated":[128],"(e.g.,":[130],"loops),":[132],"combats":[134],"\"lost-in-the-middle\"":[136],"problem":[137],"by":[138,152],"isolating":[139],"distinct":[140],"sub-prompts":[141],"(instructions,":[142],"formats,":[143],"or":[144],"few-shot":[145],"examples).":[146],"It":[147],"further":[148],"boosts":[149],"training":[150,184],"efficiency":[151],"focusing":[153],"on":[154],"error-prone":[155],"samples":[156],"through":[157,190],"selective":[158],"gradient":[159,178],"computation.":[160],"Across":[161],"diverse":[162],"tasks,":[163],"including":[164],"single-step":[165],"classification,":[166],"retrieval-based":[168],"QA,":[169],"agent-driven":[171],"pipelines,":[172],"consistently":[174],"outperforms":[175],"existing":[176],"baselines":[179],"both":[181],"accuracy":[182],"cost.":[185],"By":[186],"unifying":[187],"optimization":[189],"graph-centric":[192],"lens,":[193],"offers":[195],"powerful":[197],"new":[198],"paradigm":[199],"scaling":[201],"automating":[203],"workflows":[205],"-":[206],"mirroring":[207],"transformative":[209],"role":[210],"automatic":[212],"differentiation":[213],"libraries":[214],"long":[216],"played":[217],"neural":[219],"network":[220],"research.":[221]},"counts_by_year":[{"year":2026,"cited_by_count":22},{"year":2025,"cited_by_count":4}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2025-01-30T00:00:00"}