{"id":"https://openalex.org/W7135044728","doi":"https://doi.org/10.48550/arxiv.2603.10088","title":"ES-dLLM: Efficient Inference for Diffusion Large Language Models by Early-Skipping","display_name":"ES-dLLM: Efficient Inference for Diffusion Large Language Models by Early-Skipping","publication_year":2026,"publication_date":"2026-03-10","ids":{"openalex":"https://openalex.org/W7135044728","doi":"https://doi.org/10.48550/arxiv.2603.10088"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.10088","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10088","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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.10088","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5025588570","display_name":"Zijian Zhu","orcid":"https://orcid.org/0000-0001-8343-4936"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Zijian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128812527","display_name":"Fei Ren","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ren, Fei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083108300","display_name":"Zhanhong Tan","orcid":"https://orcid.org/0000-0001-6725-3211"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tan, Zhanhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5006570986","display_name":"Kaisheng Ma","orcid":"https://orcid.org/0000-0001-9226-3366"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Kaisheng","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":0.430400013923645,"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":0.430400013923645,"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/T13910","display_name":"Computational and Text Analysis Methods","score":0.09920000284910202,"subfield":{"id":"https://openalex.org/subfields/3300","display_name":"General Social Sciences"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.07970000058412552,"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/inference","display_name":"Inference","score":0.7742999792098999},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6866999864578247},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.670799970626831},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.6123999953269958},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5662999749183655},{"id":"https://openalex.org/keywords/autoregressive-model","display_name":"Autoregressive model","score":0.527400016784668},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4952000081539154},{"id":"https://openalex.org/keywords/acceleration","display_name":"Acceleration","score":0.4117000102996826}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8299999833106995},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7742999792098999},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6866999864578247},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.670799970626831},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6123999953269958},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5662999749183655},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.527400016784668},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4952000081539154},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.4117000102996826},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4041999876499176},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.3292999863624573},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32919999957084656},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.3190000057220459},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.3174999952316284},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.3163999915122986},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.29190000891685486},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.2824999988079071},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.2718000113964081},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.26899999380111694},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.2628999948501587}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.10088","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10088","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.10088","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10088","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":"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"large":[1],"language":[2],"models":[3,13],"(dLLMs)":[4],"are":[5],"emerging":[6],"as":[7,37],"a":[8,80],"promising":[9],"alternative":[10],"to":[11,16,19,127,143,152],"autoregressive":[12],"(ARMs)":[14],"due":[15],"their":[17],"ability":[18],"capture":[20],"bidirectional":[21],"context":[22,41],"and":[23,57,65,109,118,129,150],"the":[24,30,38,52,98,147,155],"potential":[25],"for":[26,85],"parallel":[27],"generation.":[28],"Despite":[29],"advantages,":[31],"dLLM":[32,86],"inference":[33,82],"remains":[34],"computationally":[35],"expensive":[36],"full":[39],"input":[40],"is":[42,103],"processed":[43],"at":[44],"every":[45],"iteration.":[46],"In":[47],"this":[48,75],"work,":[49],"we":[50,77],"analyze":[51],"generation":[53,161],"dynamics":[54],"of":[55,112,125],"dLLMs":[56],"find":[58],"that":[59,87,121],"intermediate":[60,106],"representations,":[61],"including":[62],"key,":[63],"value,":[64],"hidden":[66],"states,":[67],"change":[68],"only":[69],"subtly":[70],"across":[71],"successive":[72],"iterations.":[73,114],"Leveraging":[74],"insight,":[76],"propose":[78],"\\textbf{ES-dLLM},":[79],"training-free":[81],"acceleration":[83],"framework":[84],"reduces":[88],"computation":[89],"by":[90],"skipping":[91],"tokens":[92,131],"in":[93],"early":[94],"layers":[95],"based":[96],"on":[97,116,136],"estimated":[99],"importance.":[100],"Token":[101],"importance":[102],"computed":[104],"with":[105],"tensor":[107],"variation":[108],"confidence":[110],"scores":[111],"previous":[113],"Experiments":[115],"LLaDA-8B":[117],"Dream-7B":[119],"demonstrate":[120],"ES-dLLM":[122],"achieves":[123],"throughput":[124],"up":[126,151],"226.57":[128],"308.51":[130],"per":[132],"second":[133],"(TPS),":[134],"respectively,":[135],"an":[137],"NVIDIA":[138],"H200":[139],"GPU,":[140],"delivering":[141],"5.6$\\times$":[142],"16.8$\\times$":[144],"speedup":[145],"over":[146,154],"vanilla":[148],"implementation":[149],"1.85$\\times$":[153],"state-of-the-art":[156],"caching":[157],"method,":[158],"while":[159],"preserving":[160],"quality.":[162]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-03-13T00:00:00"}