{"id":"https://openalex.org/W7158049165","doi":"https://doi.org/10.48550/arxiv.2604.24927","title":"Large Language Models Explore by Latent Distilling","display_name":"Large Language Models Explore by Latent Distilling","publication_year":2026,"publication_date":"2026-04-27","ids":{"openalex":"https://openalex.org/W7158049165","doi":"https://doi.org/10.48550/arxiv.2604.24927"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.24927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.24927","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.2604.24927","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5094347633","display_name":"Yuanhao Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Yuanhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134823695","display_name":"Ao Lu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Ao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134838238","display_name":"Lufei Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Lufei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134864301","display_name":"Zheng Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134842607","display_name":"Yexin Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yexin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134823498","display_name":"Kan Ren","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ren, Kan","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":0.39160001277923584,"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":0.39160001277923584,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.09880000352859497,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13629","display_name":"Text Readability and Simplification","score":0.0674000009894371,"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/decoding-methods","display_name":"Decoding methods","score":0.5174999833106995},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.49540001153945923},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.4575999975204468},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4415000081062317},{"id":"https://openalex.org/keywords/heuristic","display_name":"Heuristic","score":0.42719998955726624},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.4169999957084656},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.3910999894142151},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.39010000228881836},{"id":"https://openalex.org/keywords/coherence","display_name":"Coherence (philosophical gambling strategy)","score":0.3675000071525574}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7530999779701233},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5620999932289124},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5174999833106995},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.49540001153945923},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.4575999975204468},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4415000081062317},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4332999885082245},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.42719998955726624},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.4169999957084656},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3910999894142151},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.39010000228881836},{"id":"https://openalex.org/C2781181686","wikidata":"https://www.wikidata.org/wiki/Q4226068","display_name":"Coherence (philosophical gambling strategy)","level":2,"score":0.3675000071525574},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.365200012922287},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3540000021457672},{"id":"https://openalex.org/C2778738651","wikidata":"https://www.wikidata.org/wiki/Q16546687","display_name":"Novelty","level":2,"score":0.3303999900817871},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3237999975681305},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3095000088214874},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.3043000102043152},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2897999882698059},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.2874999940395355},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2865999937057495},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2827000021934509},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.28130000829696655},{"id":"https://openalex.org/C19768560","wikidata":"https://www.wikidata.org/wiki/Q320727","display_name":"Dependency (UML)","level":2,"score":0.28040000796318054},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.2766999900341034},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.26739999651908875},{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.266400009393692},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.25450000166893005},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.25220000743865967},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.25209999084472656}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.24927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.24927","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.2604.24927","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.24927","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":0.8864203691482544,"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":{"Generating":[0],"diverse":[1],"responses":[2],"is":[3,44,146],"crucial":[4],"for":[5],"test-time":[6],"scaling":[7],"of":[8,90,175],"large":[9],"language":[10],"models":[11],"(LLMs),":[12],"yet":[13],"standard":[14],"stochastic":[15,185],"sampling":[16],"mostly":[17],"yields":[18],"surface-level":[19],"lexical":[20],"variation,":[21],"limiting":[22],"semantic":[23,39,143],"exploration.":[24],"In":[25],"this":[26,75],"paper,":[27],"we":[28,77],"propose":[29],"Exploratory":[30],"Sampling":[31],"(ESamp),":[32],"a":[33,79,125],"decoding":[34,140],"approach":[35],"that":[36,50,168],"explicitly":[37],"encourages":[38],"diversity":[40,206],"during":[41],"generation.":[42],"ESamp":[43,119,145,169,190],"motivated":[45],"by":[46,114],"the":[47,91,99,106,111,115,121,135,162,172,203],"well-known":[48],"observation":[49],"neural":[51],"networks":[52],"tend":[53],"to":[54,61,85,97,110,128,183],"make":[55],"lower-error":[56],"predictions":[57],"on":[58,70,74,134],"inputs":[59],"similar":[60],"those":[62],"encountered":[63],"before,":[64],"and":[65,186,197,201,207],"incur":[66],"higher":[67],"prediction":[68,122],"error":[69,123],"novel":[71],"ones.":[72],"Building":[73],"property,":[76],"train":[78],"lightweight":[80],"Distiller":[81,107],"at":[82],"test":[83],"time":[84],"predict":[86],"deep-layer":[87],"hidden":[88],"representations":[89,96],"LLM":[92],"from":[93],"its":[94],"shallow-layer":[95],"model":[98],"LLM's":[100],"depth-wise":[101],"representation":[102],"transitions.":[103],"During":[104],"decoding,":[105],"continuously":[108],"adapts":[109],"mappings":[112],"induced":[113],"current":[116,136],"generation":[117,199],"context.":[118],"uses":[120],"as":[124],"novelty":[126],"signal":[127],"reweight":[129],"candidate":[130],"token":[131],"extensions":[132],"conditioned":[133],"prefix,":[137],"thereby":[138],"biasing":[139],"toward":[141],"less-explored":[142],"patterns.":[144],"implemented":[147],"with":[148,153],"an":[149],"asynchronous":[150],"training--inference":[151],"pipeline,":[152],"less":[154],"than":[155],"5%":[156],"worst":[157],"case":[158],"overhead":[159],"(1.2%":[160],"in":[161,209],"optimized":[163],"release).":[164],"Empirical":[165],"results":[166],"show":[167],"significantly":[170],"boosts":[171],"Pass@k":[173],"efficiency":[174],"reasoning":[176],"models,":[177],"showing":[178],"superior":[179],"or":[180],"comparable":[181],"performance":[182],"strong":[184],"heuristic":[187],"baselines.":[188],"Notably,":[189],"achieves":[191],"robust":[192],"generalization":[193],"across":[194],"mathematics,":[195],"science,":[196],"code":[198,213],"benchmarks":[200],"breaks":[202],"trade-off":[204],"between":[205],"coherence":[208],"creative":[210],"writing.":[211],"Our":[212],"has":[214],"released":[215],"at:":[216],"https://github.com/LinesHogan/tLLM.":[217]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-30T00:00:00"}