{"id":"https://openalex.org/W7127391691","doi":"https://doi.org/10.48550/arxiv.2602.00397","title":"Fast Forward: Accelerating LLM Prefill with Predictive FFN Sparsity","display_name":"Fast Forward: Accelerating LLM Prefill with Predictive FFN Sparsity","publication_year":2026,"publication_date":"2026-01-30","ids":{"openalex":"https://openalex.org/W7127391691","doi":"https://doi.org/10.48550/arxiv.2602.00397"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.00397","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124969514","display_name":"Aayush Gautam","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Gautam, Aayush","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018463719","display_name":"Mukul Gagrani","orcid":"https://orcid.org/0000-0003-0483-1090"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gagrani, Mukul","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124950072","display_name":"Junyoung Park","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Park, Junyoung","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124901745","display_name":"Mingu Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Mingu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124895572","display_name":"Chiris Lott","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lott, Chiris","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5124970607","display_name":"Narasimha Reddy","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Reddy, Narasimha","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5124969514"],"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/T10036","display_name":"Advanced Neural Network Applications","score":0.3327000141143799,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.3327000141143799,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.06449999660253525,"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":0.040300000458955765,"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/inference","display_name":"Inference","score":0.6557999849319458},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.6168000102043152},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.5781000256538391},{"id":"https://openalex.org/keywords/autoregressive-model","display_name":"Autoregressive model","score":0.5699999928474426},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.557200014591217},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5242000222206116},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.4392000138759613}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7462000250816345},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6557999849319458},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6168000102043152},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5781000256538391},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.5699999928474426},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.557200014591217},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5242000222206116},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4392000138759613},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39959999918937683},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.38359999656677246},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3813000023365021},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.358599990606308},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.35350000858306885},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.3111000061035156},{"id":"https://openalex.org/C917703","wikidata":"https://www.wikidata.org/wiki/Q7239668","display_name":"Predictive inference","level":5,"score":0.29910001158714294},{"id":"https://openalex.org/C2780023022","wikidata":"https://www.wikidata.org/wiki/Q1338171","display_name":"Compensation (psychology)","level":2,"score":0.28999999165534973},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.28859999775886536},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.25380000472068787},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.00397","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},{"id":"doi:10.48550/arxiv.2602.00397","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.00397","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"article"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.00397","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"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":{"The":[0],"prefill":[1,48,68],"stage":[2],"of":[3,32],"large":[4],"language":[5],"model":[6],"(LLM)":[7],"inference":[8,149],"is":[9],"a":[10,61,77,98],"key":[11],"computational":[12],"bottleneck":[13],"for":[14,30,41,145],"long-context":[15,147],"workloads.":[16],"At":[17],"short-to-moderate":[18],"context":[19],"lengths":[20],"(1K--16K":[21],"tokens),":[22],"Feed-Forward":[23],"Networks":[24],"(FFNs)":[25],"dominate":[26],"this":[27],"cost,":[28],"accounting":[29],"most":[31],"the":[33,47,136],"total":[34],"FLOPs.":[35],"Existing":[36],"FFN":[37,72,127],"sparsification":[38],"methods,":[39],"designed":[40],"autoregressive":[42],"decoding,":[43],"fail":[44],"to":[45,81,92,102,115,121,135],"exploit":[46],"stage's":[49],"parallelism":[50],"and":[51,96,111],"often":[52],"degrade":[53],"accuracy.":[54],"To":[55],"address":[56],"this,":[57],"we":[58],"introduce":[59],"FastForward,":[60],"predictive":[62],"sparsity":[63,100,128],"framework":[64],"that":[65],"accelerates":[66],"LLM":[67,148],"through":[69],"block-wise,":[70],"context-aware":[71],"sparsity.":[73],"FastForward":[74,118],"combines":[75],"(1)":[76],"lightweight":[78],"expert":[79],"predictor":[80],"select":[82],"high-importance":[83],"neurons":[84],"per":[85],"block,":[86],"(2)":[87],"an":[88],"error":[89],"compensation":[90],"network":[91],"correct":[93],"sparsity-induced":[94],"errors,":[95],"(3)":[97],"layer-wise":[99],"scheduler":[101],"allocate":[103],"compute":[104],"based":[105],"on":[106,139,150],"token-mixing":[107],"importance.":[108],"Across":[109],"LLaMA":[110],"Qwen":[112],"models":[113],"up":[114,120],"8B":[116],"parameters,":[117],"delivers":[119],"1.45$\\times$":[122],"compute-bound":[123],"speedup":[124],"at":[125],"50%":[126],"with":[129],"$<$":[130],"6%":[131],"accuracy":[132],"loss":[133],"compared":[134],"dense":[137],"baseline":[138],"LongBench,":[140],"substantially":[141],"reducing":[142],"Time-to-First-Token":[143],"(TTFT)":[144],"efficient,":[146],"constrained":[151],"hardware.":[152]},"counts_by_year":[],"updated_date":"2026-04-04T16:13:02.066488","created_date":"2026-02-04T00:00:00"}