{"id":"https://openalex.org/W4407184045","doi":"https://doi.org/10.48550/arxiv.2502.01567","title":"Latent Thought Models with Variational Bayes Inference-Time Computation","display_name":"Latent Thought Models with Variational Bayes Inference-Time Computation","publication_year":2025,"publication_date":"2025-02-03","ids":{"openalex":"https://openalex.org/W4407184045","doi":"https://doi.org/10.48550/arxiv.2502.01567"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2502.01567","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.01567","pdf_url":"https://arxiv.org/pdf/2502.01567","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/2502.01567","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Kong, Deqian","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kong, Deqian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101414990","display_name":"Minglu Zhao","orcid":"https://orcid.org/0009-0005-3632-4167","observed_orcids":["https://orcid.org/0009-0005-3632-4167"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Minglu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5099054886","display_name":"Dehong Xu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Dehong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101799605","display_name":"Bo Pang","orcid":"https://orcid.org/0000-0003-4521-6369","observed_orcids":["https://orcid.org/0000-0003-4521-6369"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pang, Bo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041713522","display_name":"Shu Wang","orcid":"https://orcid.org/0009-0001-7554-1835","observed_orcids":["https://orcid.org/0009-0001-7554-1835"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Shu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Honig, Edouardo","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Honig, Edouardo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013974468","display_name":"Zhangzhang Si","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Si, Zhangzhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037372838","display_name":"Chuan Li","orcid":"https://orcid.org/0009-0004-8041-2538","observed_orcids":["https://orcid.org/0009-0004-8041-2538"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Chuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104112430","display_name":"Jianwen Xie","orcid":"https://orcid.org/0009-0008-9271-4022","observed_orcids":["https://orcid.org/0009-0008-9271-4022"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Jianwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003810400","display_name":"Sirui Xie","orcid":"https://orcid.org/0000-0002-0295-2588","observed_orcids":["https://orcid.org/0000-0002-0295-2588"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Sirui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5101780958","display_name":"Ying Nian Wu","orcid":"https://orcid.org/0009-0001-6768-5118","observed_orcids":["https://orcid.org/0009-0001-6768-5118","https://orcid.org/0009-0002-4728-1279","https://orcid.org/0000-0002-3778-8183"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Ying Nian","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":true,"cited_by_count":24,"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":0.9564999938011169,"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":0.9564999938011169,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.0019000000320374966,"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/variational-bayesian-inference","display_name":"variational inference","score":0.6340000033378601},{"id":"https://openalex.org/keywords/sample-efficiency","display_name":"sample efficiency","score":0.6169999837875366},{"id":"https://openalex.org/keywords/parameter-efficiency","display_name":"parameter efficiency","score":0.9580000042915344},{"id":"https://openalex.org/keywords/text-generation","display_name":"text generation","score":0.7300000190734863},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.640999972820282}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.736868143081665},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5869746804237366},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.579572856426239},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4501476585865021},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.43721163272857666},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.41556647419929504},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.05797320604324341}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2502.01567","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.01567","pdf_url":"https://arxiv.org/pdf/2502.01567","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.2502.01567","doi":"https://doi.org/10.48550/arxiv.2502.01567","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2502.01567","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:2502.01567","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2502.01567","pdf_url":"https://arxiv.org/pdf/2502.01567","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":[{"id":"https://openalex.org/G1038160048","display_name":"Generative Modeling with Short Run Computing","funder_award_id":"2015577","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G6791872014","display_name":"Collaborative Research: Graph Learning, Generation, and Optimization with Highly Expressive Graph Neural Network Models","funder_award_id":"2415226","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"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4407184045.pdf","grobid_xml":"https://content.openalex.org/works/W4407184045.grobid-xml"},"referenced_works_count":26,"referenced_works":["https://openalex.org/W1539309091","https://openalex.org/W1959608418","https://openalex.org/W2325237720","https://openalex.org/W2473344385","https://openalex.org/W2525332836","https://openalex.org/W2587741066","https://openalex.org/W2798240283","https://openalex.org/W2908510526","https://openalex.org/W2938704169","https://openalex.org/W2964121744","https://openalex.org/W3001279689","https://openalex.org/W3210904070","https://openalex.org/W4224308101","https://openalex.org/W4282958700","https://openalex.org/W4307207888","https://openalex.org/W4310421098","https://openalex.org/W4379470463","https://openalex.org/W4385571393","https://openalex.org/W4403573100","https://openalex.org/W4403751234","https://openalex.org/W4405254469","https://openalex.org/W4405354515","https://openalex.org/W4414281281","https://openalex.org/W4415796024","https://openalex.org/W4415800813","https://openalex.org/W7133225605"],"related_works":["https://openalex.org/W7223019693","https://openalex.org/W4415329948","https://openalex.org/W7083813732","https://openalex.org/W4404317990","https://openalex.org/W7151608107","https://openalex.org/W7164921006","https://openalex.org/W7127567150","https://openalex.org/W4415250069","https://openalex.org/W4414804979","https://openalex.org/W4417298486"],"abstract_inverted_index":{"We":[0],"propose":[1],"a":[2,39,44],"novel":[3],"class":[4],"of":[5,35,56,64,72,95,102],"language":[6,161],"models,":[7],"Latent":[8],"Thought":[9],"Models":[10,89],"(LTMs),":[11],"which":[12],"incorporate":[13],"explicit":[14,21],"latent":[15,25,28,65,103],"thought":[16,29,104],"vectors":[17,30,66],"that":[18,79,172],"follow":[19],"an":[20],"prior":[22],"model":[23,124,175],"in":[24,97,156,181],"space.":[26],"These":[27],"guide":[31],"the":[32,49,61,93],"autoregressive":[33,145],"generation":[34],"ground":[36],"tokens":[37],"through":[38],"Transformer":[40],"decoder.":[41],"Training":[42],"employs":[43],"dual-rate":[45],"optimization":[46],"process":[47],"within":[48],"classical":[50],"variational":[51,58],"Bayes":[52],"framework:":[53],"fast":[54],"learning":[55,71],"local":[57],"parameters":[59],"for":[60,126],"posterior":[62],"distribution":[63],"(inference-time":[67],"computation),":[68],"and":[69,100,140,147,159,177,183],"slow":[70],"global":[73],"decoder":[74],"parameters.":[75],"Empirical":[76],"studies":[77],"reveal":[78],"LTMs":[80,136,165],"possess":[81],"additional":[82],"scaling":[83,134],"dimensions":[84],"beyond":[85],"traditional":[86],"Large":[87],"Language":[88],"(LLMs),":[90],"such":[91],"as":[92],"number":[94,101],"iterations":[96],"inference-time":[98],"computation":[99],"vectors.":[105],"Higher":[106],"sample":[107,139],"efficiency":[108,142],"can":[109],"be":[110],"achieved":[111],"by":[112,122],"increasing":[113],"training":[114],"compute":[115],"per":[116],"token,":[117],"with":[118,174],"further":[119],"gains":[120],"possible":[121],"trading":[123],"size":[125],"more":[127],"inference":[128],"steps.":[129],"Designed":[130],"based":[131],"on":[132],"these":[133,154],"properties,":[135],"demonstrate":[137],"superior":[138],"parameter":[141],"compared":[143],"to":[144],"models":[146],"discrete":[148],"diffusion":[149],"models.":[150],"They":[151],"significantly":[152],"outperform":[153],"counterparts":[155],"validation":[157],"perplexity":[158],"zero-shot":[160],"modeling":[162],"tasks.":[163],"Additionally,":[164],"exhibit":[166],"emergent":[167],"few-shot":[168],"in-context":[169],"reasoning":[170],"capabilities":[171],"scale":[173],"size,":[176],"achieve":[178],"competitive":[179],"performance":[180],"conditional":[182],"unconditional":[184],"text":[185],"generation.":[186]},"counts_by_year":[{"year":2026,"cited_by_count":12},{"year":2025,"cited_by_count":12}],"updated_date":"2026-11-11T12:16:26.678958","created_date":"2025-02-06T00:00:00"}