{"id":"https://openalex.org/W7161059234","doi":"https://doi.org/10.48550/arxiv.2605.11011","title":"LoopUS: Recasting Pretrained LLMs into Looped Latent Refinement Models","display_name":"LoopUS: Recasting Pretrained LLMs into Looped Latent Refinement Models","publication_year":2026,"publication_date":"2026-05-10","ids":{"openalex":"https://openalex.org/W7161059234","doi":"https://doi.org/10.48550/arxiv.2605.11011"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.11011","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11011","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.2605.11011","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126822074","display_name":"Taekhyun Park","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Park, Taekhyun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136014177","display_name":"Yongjae Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Yongjae","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136013316","display_name":"Dohee Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Dohee","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5047158713","display_name":"Hyerim Bae","orcid":"https://orcid.org/0000-0003-2602-5911"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bae, Hyerim","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":0.22779999673366547,"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":0.22779999673366547,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.06949999928474426,"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"}},{"id":"https://openalex.org/T11450","display_name":"Model-Driven Software Engineering Techniques","score":0.05909999832510948,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/representation","display_name":"Representation (politics)","score":0.6809999942779541},{"id":"https://openalex.org/keywords/block","display_name":"Block (permutation group theory)","score":0.5644999742507935},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.550599992275238},{"id":"https://openalex.org/keywords/compromise","display_name":"Compromise","score":0.4059000015258789},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.3953000009059906},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.3919999897480011},{"id":"https://openalex.org/keywords/core","display_name":"Core (optical fiber)","score":0.34130001068115234},{"id":"https://openalex.org/keywords/feature-learning","display_name":"Feature learning","score":0.3407000005245209}],"concepts":[{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.6809999942779541},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6794000267982483},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6189000010490417},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.5644999742507935},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.550599992275238},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5349000096321106},{"id":"https://openalex.org/C46355384","wikidata":"https://www.wikidata.org/wiki/Q726686","display_name":"Compromise","level":2,"score":0.4059000015258789},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3953000009059906},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.3919999897480011},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.34130001068115234},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.3407000005245209},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.334199994802475},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.31540000438690186},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C2777601897","wikidata":"https://www.wikidata.org/wiki/Q3409113","display_name":"Presentation (obstetrics)","level":2,"score":0.2833999991416931},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.2741999924182892},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.2667999863624573},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.25920000672340393},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.2547999918460846},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.250900000333786}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.11011","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11011","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.2605.11011","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11011","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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Looped":[0],"computation":[1],"shows":[2],"promise":[3],"in":[4],"improving":[5],"the":[6,70,161],"reasoning-oriented":[7,157],"performance":[8,158],"of":[9],"LLMs":[10],"by":[11,96],"scaling":[12],"test-time":[13],"compute.":[14],"However,":[15],"existing":[16],"approaches":[17],"typically":[18],"require":[19],"either":[20],"training":[21,167],"recurrent":[22,166],"models":[23],"from":[24,168],"scratch":[25],"or":[26,164],"applying":[27],"disruptive":[28],"retrofits,":[29],"which":[30],"involve":[31],"substantial":[32],"computational":[33,146],"costs":[34],"and":[35,80,120,148],"may":[36],"compromise":[37],"pretrained":[38,57,71],"capabilities.":[39],"To":[40],"address":[41],"these":[42,130],"limitations,":[43],"we":[44],"introduce":[45],"\\textbf{Looped":[46],"Depth":[47],"Up-Scaling}":[48],"(LoopUS),":[49],"a":[50,55,60,64,76,81,122,133,138],"post-training":[51],"framework":[52],"that":[53],"converts":[54],"standard":[56,134],"LLM":[58,72],"into":[59,73,137],"looped":[61,77,139],"architecture.":[62],"As":[63],"key":[65],"technical":[66],"contribution,":[67],"LoopUS":[68,155],"recasts":[69],"an":[74,101],"encoder,":[75],"reasoning":[78],"block,":[79],"decoder.":[82],"It":[83],"operationalizes":[84],"this":[85],"latent-refinement":[86],"architecture":[87],"through":[88],"four":[89],"core":[90],"components:":[91],"(1)":[92],"block":[93],"decomposition,":[94],"guided":[95],"staged":[97],"representation":[98,149],"dynamics;":[99],"(2)":[100],"input-dependent":[102],"selective":[103],"gate":[104],"to":[105],"mitigate":[106],"hidden-state":[107],"drift;":[108],"(3)":[109],"random":[110],"deep":[111],"supervision":[112],"for":[113,125],"memory-efficient":[114],"learning":[115],"over":[116],"long":[117],"recursive":[118],"horizons;":[119],"(4)":[121],"confidence":[123],"head":[124],"adaptive":[126],"early":[127],"exiting.":[128],"Collectively,":[129],"mechanisms":[131],"transform":[132],"non-looped":[135],"model":[136],"form":[140],"while":[141],"stabilizing":[142],"it":[143],"against":[144],"both":[145],"bottlenecks":[147],"collapse.":[150],"Through":[151],"stable":[152],"latent":[153],"looping,":[154],"improves":[156],"without":[159],"extending":[160],"generated":[162],"traces":[163],"requiring":[165],"scratch.":[169],"For":[170],"more":[171],"details,":[172],"see":[173],"https://thrillcrazyer.github.io/LoopUS":[174]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-14T00:00:00"}