{"id":"https://openalex.org/W7162634154","doi":"https://doi.org/10.48550/arxiv.2605.28360","title":"Prompt Codebooks: Discrete Compositional Optimization for Language Model Instruction Refinement","display_name":"Prompt Codebooks: Discrete Compositional Optimization for Language Model Instruction Refinement","publication_year":2026,"publication_date":"2026-05-27","ids":{"openalex":"https://openalex.org/W7162634154","doi":"https://doi.org/10.48550/arxiv.2605.28360"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.28360","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.28360","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,"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":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.28360","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137267354","display_name":"Jyotirmoy Nath","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nath, Jyotirmoy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137220816","display_name":"Neeraj Kumar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kumar, Neeraj","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137254781","display_name":"Brejesh Lall","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lall, Brejesh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"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":false,"cited_by_count":0,"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":0.3149000108242035,"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":0.3149000108242035,"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":0.29179999232292175,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.08640000224113464,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/generator","display_name":"Generator (circuit theory)","score":0.6007999777793884},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.5275999903678894},{"id":"https://openalex.org/keywords/codebook","display_name":"Codebook","score":0.5067999958992004},{"id":"https://openalex.org/keywords/string","display_name":"String (physics)","score":0.4860999882221222},{"id":"https://openalex.org/keywords/vagueness","display_name":"Vagueness","score":0.45509999990463257},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.4499000012874603},{"id":"https://openalex.org/keywords/structured-prediction","display_name":"Structured prediction","score":0.3950999975204468}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7315000295639038},{"id":"https://openalex.org/C2780992000","wikidata":"https://www.wikidata.org/wiki/Q17016113","display_name":"Generator (circuit theory)","level":3,"score":0.6007999777793884},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.5275999903678894},{"id":"https://openalex.org/C127759330","wikidata":"https://www.wikidata.org/wiki/Q637416","display_name":"Codebook","level":2,"score":0.5067999958992004},{"id":"https://openalex.org/C157486923","wikidata":"https://www.wikidata.org/wiki/Q1376436","display_name":"String (physics)","level":2,"score":0.4860999882221222},{"id":"https://openalex.org/C2776825360","wikidata":"https://www.wikidata.org/wiki/Q1411921","display_name":"Vagueness","level":3,"score":0.45509999990463257},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4499000012874603},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4018999934196472},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.3950999975204468},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.38089999556541443},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.36970001459121704},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.34369999170303345},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3353999853134155},{"id":"https://openalex.org/C9996903","wikidata":"https://www.wikidata.org/wiki/Q1749235","display_name":"Cryptographic nonce","level":3,"score":0.31380000710487366},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27950000762939453},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2777999937534332},{"id":"https://openalex.org/C165297611","wikidata":"https://www.wikidata.org/wiki/Q43249","display_name":"Morpheme","level":2,"score":0.2678999900817871},{"id":"https://openalex.org/C3232509","wikidata":"https://www.wikidata.org/wiki/Q18237","display_name":"Instinct","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.2596000134944916},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.2590000033378601},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.25189998745918274}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.28360","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.28360","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,"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:10.48550/arxiv.2605.28360","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.28360","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,"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":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"score":0.6558582186698914,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Automatic":[0],"prompt":[1,18,46,95,182],"optimization":[2,47],"(APO)":[3],"has":[4],"driven":[5],"significant":[6],"gains":[7],"in":[8,71,134,177],"LLM-based":[9,87],"agentic":[10],"workflows.":[11],"However,":[12],"existing":[13],"methods":[14],"treat":[15],"each":[16,77],"task's":[17],"as":[19,52],"a":[20,43,56,72,80,89,94,101,104,123,142],"monolithic,":[21],"instance-blind":[22,147],"string":[23],"optimized":[24],"through":[25],"global":[26],"edits,":[27],"producing":[28],"brittle":[29],"updates":[30],"and":[31,75,120,154,175,179,190],"preventing":[32],"the":[33,97,117,135,166],"reuse":[34],"of":[35,59,83],"learned":[36],"sub-behaviors.":[37],"We":[38],"propose":[39],"Prompt":[40],"Codebooks":[41],"(PCO),":[42],"novel":[44],"compositional":[45],"framework":[48],"that":[49,107],"recasts":[50],"APO":[51],"discrete":[53,73],"learning":[54],"over":[55,158],"finite":[57],"vocabulary":[58],"natural-language":[60],"instincts":[61],"-":[62],"atomic,":[63],"reusable":[64],"instruction":[65],"units.":[66],"PCO":[67,156],"organizes":[68],"prompt-construction":[69],"knowledge":[70],"codebook":[74,121],"routes":[76],"input":[78],"to":[79,162,186],"small":[81],"subset":[82],"entries":[84],"via":[85],"an":[86],"encoder;":[88],"generator":[90],"composes":[91],"them":[92],"into":[93,111],"for":[96],"frozen":[98],"target":[99],"model;":[100],"critic":[102],"emits":[103],"structured":[105],"verdict":[106],"decomposes":[108],"by":[109,160,171,184],"attribution":[110],"per-variable":[112],"textual":[113],"gradients,":[114],"jointly":[115],"training":[116],"encoder,":[118],"generator,":[119],"under":[122,146],"language-valued":[124],"min-max":[125],"objective.":[126],"The":[127],"resulting":[128],"routing":[129],"is":[130],"per-instance:":[131],"different":[132,139],"inputs":[133],"same":[136],"task":[137],"receive":[138],"instinct":[140],"compositions,":[141],"regime":[143],"structurally":[144],"inexpressible":[145],"methods.":[148],"Across":[149],"six":[150],"benchmarks":[151],"on":[152,173],"Qwen3-8B":[153],"LLaMA-3.1-8B,":[155],"improves":[157],"zero-shot":[159],"up":[161,185],"+30.36":[163],"points,":[164],"surpasses":[165],"strongest":[167],"prior":[168],"baseline":[169],"(GEPA)":[170],"+3.34":[172],"HotpotQA":[174],"+1.11":[176],"aggregate,":[178],"reduces":[180],"deployed":[181],"length":[183],"14.1x":[187],"versus":[188,192],"MIPROv2":[189],"3.0x":[191],"GEPA":[193],"using":[194],"only":[195],"K=16":[196],"instincts.":[197]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-29T00:00:00"}