{"id":"https://openalex.org/W4414889978","doi":"https://doi.org/10.1145/3786335.3813168","title":"Scaling Textual Gradients via Sampling-Based Momentum","display_name":"Scaling Textual Gradients via Sampling-Based Momentum","publication_year":2026,"publication_date":"2026-05-22","ids":{"openalex":"https://openalex.org/W4414889978","doi":"https://doi.org/10.1145/3786335.3813168"},"language":"en","primary_location":{"id":"doi:10.1145/3786335.3813168","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3786335.3813168","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM Conference on AI and Agentic Systems","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3786335.3813168","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Zixin Ding","orcid":"https://orcid.org/0009-0007-2227-1046"},"institutions":[{"id":"https://openalex.org/I40347166","display_name":"University of Chicago","ror":"https://ror.org/024mw5h28","country_code":"US","type":"education","lineage":["https://openalex.org/I40347166"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zixin Ding","raw_affiliation_strings":["University of Chicago, Chicago, USA"],"raw_orcid":"https://orcid.org/0009-0007-2227-1046","affiliations":[{"raw_affiliation_string":"University of Chicago, Chicago, USA","institution_ids":["https://openalex.org/I40347166"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Junyuan Hong","orcid":"https://orcid.org/0000-0002-5718-5187"},"institutions":[{"id":"https://openalex.org/I86519309","display_name":"The University of Texas at Austin","ror":"https://ror.org/00hj54h04","country_code":"US","type":"education","lineage":["https://openalex.org/I86519309"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Junyuan Hong","raw_affiliation_strings":["University of Texas at Austin, Austin, USA"],"raw_orcid":"https://orcid.org/0000-0002-5718-5187","affiliations":[{"raw_affiliation_string":"University of Texas at Austin, Austin, USA","institution_ids":["https://openalex.org/I86519309"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Zhan Shi","orcid":"https://orcid.org/0009-0002-1795-3409"},"institutions":[{"id":"https://openalex.org/I16269868","display_name":"Santa Clara University","ror":"https://ror.org/03ypqe447","country_code":"US","type":"education","lineage":["https://openalex.org/I16269868"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhan Shi","raw_affiliation_strings":["Santa Clara University, Santa Clara, USA"],"raw_orcid":"https://orcid.org/0009-0002-1795-3409","affiliations":[{"raw_affiliation_string":"Santa Clara University, Santa Clara, USA","institution_ids":["https://openalex.org/I16269868"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Tianhao Wang","orcid":"https://orcid.org/0000-0001-7944-3418"},"institutions":[{"id":"https://openalex.org/I20089843","display_name":"Princeton University","ror":"https://ror.org/00hx57361","country_code":"US","type":"education","lineage":["https://openalex.org/I20089843"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tianhao Wang","raw_affiliation_strings":["Princeton University, Princeton, USA"],"raw_orcid":"https://orcid.org/0000-0001-7944-3418","affiliations":[{"raw_affiliation_string":"Princeton University, Princeton, USA","institution_ids":["https://openalex.org/I20089843"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101422411","display_name":"Zinan Lin","orcid":"https://orcid.org/0000-0002-8421-2662"},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zinan Lin","raw_affiliation_strings":["Microsoft Research, Redmond, USA"],"raw_orcid":"https://orcid.org/0000-0002-8421-2662","affiliations":[{"raw_affiliation_string":"Microsoft Research, Redmond, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Li Yin","orcid":"https://orcid.org/0000-0003-1063-0682"},"institutions":[{"id":"https://openalex.org/I3080301080","display_name":"San Francisco Symphony","ror":"https://ror.org/054ff8869","country_code":"US","type":"other","lineage":["https://openalex.org/I3080301080"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Li Yin","raw_affiliation_strings":["SylphAI, San Francisco, USA"],"raw_orcid":"https://orcid.org/0000-0003-1063-0682","affiliations":[{"raw_affiliation_string":"SylphAI, San Francisco, USA","institution_ids":["https://openalex.org/I3080301080"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Meng Liu","orcid":"https://orcid.org/0000-0002-9420-3874"},"institutions":[{"id":"https://openalex.org/I3080301080","display_name":"San Francisco Symphony","ror":"https://ror.org/054ff8869","country_code":"US","type":"other","lineage":["https://openalex.org/I3080301080"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Meng Liu","raw_affiliation_strings":["SylphAI, San Francisco, USA"],"raw_orcid":"https://orcid.org/0000-0002-9420-3874","affiliations":[{"raw_affiliation_string":"SylphAI, San Francisco, USA","institution_ids":["https://openalex.org/I3080301080"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048522863","display_name":"Zhangyang Wang","orcid":"https://orcid.org/0000-0002-2050-5693"},"institutions":[{"id":"https://openalex.org/I86519309","display_name":"The University of Texas at Austin","ror":"https://ror.org/00hj54h04","country_code":"US","type":"education","lineage":["https://openalex.org/I86519309"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhangyang Wang","raw_affiliation_strings":["University of Texas at Austin, Austin, USA"],"raw_orcid":"https://orcid.org/0000-0002-2050-5693","affiliations":[{"raw_affiliation_string":"University of Texas at Austin, Austin, USA","institution_ids":["https://openalex.org/I86519309"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100416071","display_name":"Yuxin Chen","orcid":"https://orcid.org/0000-0001-5955-687X"},"institutions":[{"id":"https://openalex.org/I40347166","display_name":"University of Chicago","ror":"https://ror.org/024mw5h28","country_code":"US","type":"education","lineage":["https://openalex.org/I40347166"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuxin Chen","raw_affiliation_strings":["University of Chicago, Chicago, USA"],"raw_orcid":"https://orcid.org/0000-0003-2133-140X","affiliations":[{"raw_affiliation_string":"University of Chicago, Chicago, USA","institution_ids":["https://openalex.org/I40347166"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"917","last_page":"952"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.8180999755859375,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.8180999755859375,"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/T10028","display_name":"Topic Modeling","score":0.7893000245094299,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.7598000168800354,"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/scaling","display_name":"Scaling","score":0.805400013923645},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.6902999877929688},{"id":"https://openalex.org/keywords/stochastic-gradient-descent","display_name":"Stochastic gradient descent","score":0.6431999802589417},{"id":"https://openalex.org/keywords/momentum","display_name":"Momentum (technical analysis)","score":0.6233999729156494},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.6121000051498413},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.5552999973297119},{"id":"https://openalex.org/keywords/stability","display_name":"Stability (learning theory)","score":0.5543000102043152}],"concepts":[{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.805400013923645},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.6902999877929688},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.6431999802589417},{"id":"https://openalex.org/C60718061","wikidata":"https://www.wikidata.org/wiki/Q1414747","display_name":"Momentum (technical analysis)","level":2,"score":0.6233999729156494},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6121000051498413},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.60589998960495},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.5552999973297119},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.5543000102043152},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.513700008392334},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4487000107765198},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3901999890804291},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.37700000405311584},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.31450000405311584},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.30970001220703125},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3077000081539154},{"id":"https://openalex.org/C8272713","wikidata":"https://www.wikidata.org/wiki/Q176737","display_name":"Stochastic process","level":2,"score":0.3050999939441681},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.296099990606308},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.29010000824928284},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.2761000096797943},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.27000001072883606},{"id":"https://openalex.org/C194387892","wikidata":"https://www.wikidata.org/wiki/Q1747770","display_name":"Stochastic optimization","level":2,"score":0.2680000066757202},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2581000030040741},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.25760000944137573},{"id":"https://openalex.org/C21080849","wikidata":"https://www.wikidata.org/wiki/Q13611879","display_name":"Data point","level":2,"score":0.2572999894618988}],"mesh":[],"locations_count":4,"locations":[{"id":"doi:10.1145/3786335.3813168","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3786335.3813168","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM Conference on AI and Agentic Systems","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2506.00400","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2506.00400","pdf_url":"https://arxiv.org/pdf/2506.00400","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"pmh:oai:arXiv.org:2506.00400","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2506.00400","pdf_url":"https://arxiv.org/pdf/2506.00400","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2506.00400","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2506.00400","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.1145/3786335.3813168","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3786335.3813168","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM Conference on AI and Agentic Systems","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2540632084","display_name":null,"funder_award_id":"2037026","funder_id":"https://openalex.org/F4320337391","funder_display_name":"Division of Civil, Mechanical and Manufacturing Innovation"},{"id":"https://openalex.org/G5317869797","display_name":"FMRG: Manufacturing ADvanced Electronics through Printing Using Bio-based and Locally Identifiable Compounds (MADE-PUBLIC)","funder_award_id":"2037026","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G8985754690","display_name":"Collaborative Research: RI: Small: Foundations of Few-Round Active Learning","funder_award_id":"2313131","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320337391","display_name":"Division of Civil, Mechanical and Manufacturing Innovation","ror":"https://ror.org/028yd4c30"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"LLM-based":[0],"prompt":[1,20,145],"optimization,":[2],"which":[3,93],"uses":[4],"LLM-provided":[5],"\u201ctextual":[6],"gradients\u201d":[7],"(feedback)":[8],"to":[9,60],"refine":[10],"prompts,":[11],"has":[12],"emerged":[13],"as":[14,104],"an":[15,66],"effective":[16,115],"method":[17],"for":[18],"automatic":[19],"engineering.":[21],"However,":[22],"its":[23],"scalability":[24],"and":[25,40,65,113,151,153],"stability":[26],"are":[27],"unclear":[28],"when":[29],"using":[30,99],"more":[31],"data":[32,45],"in":[33,46,80],"training.":[34],"We":[35,50],"systematically":[36],"investigate":[37],"the":[38,130],"potential":[39],"challenges":[41],"of":[42],"scaling":[43,54,116],"training":[44,55],"textual":[47],"gradient":[48,82],"descent.":[49],"show":[51],"that":[52],"naively":[53],"examples":[56],"is":[57],"infeasible":[58],"due":[59],"both":[61],"explicit":[62],"context-length":[63],"limits":[64],"implicit":[67],"context":[68,120,138],"wall,":[69],"where":[70],"long-context":[71],"degradation":[72],"yields":[73],"diminishing":[74],"returns.":[75],"Inspired":[76],"by":[77,127],"prior":[78,124],"wisdom":[79],"stochastic":[81],"descent,":[83],"we":[84],"propose":[85],"Textual":[86],"Stochastic":[87],"Gradient":[88],"Descent":[89],"with":[90],"Momentum":[91],"(TSGD-M),":[92],"reweights":[94],"updates":[95],"through":[96],"momentum":[97],"sampling,":[98],"bootstrapped":[100],"minibatch":[101],"validation":[102],"accuracy":[103],"importance":[105],"weights":[106],"over":[107],"historical":[108],"prompts.":[109],"To":[110],"stabilize":[111],"TSGD":[112],"enable":[114],"within":[117],"a":[118],"limited":[119],"window,":[121],"TSGD-M":[122,140],"carries":[123],"prompts":[125,134],"information":[126],"dynamically":[128],"exploring":[129],"past":[131],"top":[132],"performing":[133],"without":[135],"expanding":[136],"input":[137],"length.":[139],"integrates":[141],"seamlessly":[142],"into":[143],"existing":[144],"optimization":[146],"frameworks,":[147],"including":[148],"TextGrad,":[149],"DSPy-COPRO,":[150],"AdalFlow,":[152],"achieves":[154],"consistent":[155],"gains":[156],"across":[157],"6":[158],"benchmarks.":[159]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-07T00:00:00"}