{"id":"https://openalex.org/W7162282921","doi":"https://doi.org/10.48550/arxiv.2605.23171","title":"Understanding and Improving Noisy Embedding Techniques in Instruction Finetuning","display_name":"Understanding and Improving Noisy Embedding Techniques in Instruction Finetuning","publication_year":2026,"publication_date":"2026-05-22","ids":{"openalex":"https://openalex.org/W7162282921","doi":"https://doi.org/10.48550/arxiv.2605.23171"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.23171","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23171","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":null,"license_id":null,"version":null,"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":"https://doi.org/10.48550/arxiv.2605.23171","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136897712","display_name":"Abhay Yadav","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Yadav, Abhay","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5136897712"],"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":0.2563999891281128,"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":0.2563999891281128,"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":0.09570000320672989,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.06780000030994415,"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/noise","display_name":"Noise (video)","score":0.6432999968528748},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.5461999773979187},{"id":"https://openalex.org/keywords/function","display_name":"Function (biology)","score":0.45730000734329224},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4562000036239624},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.4318000078201294},{"id":"https://openalex.org/keywords/gaussian-noise","display_name":"Gaussian noise","score":0.39340001344680786},{"id":"https://openalex.org/keywords/empirical-research","display_name":"Empirical research","score":0.3864000141620636}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7472000122070312},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.6432999968528748},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.5461999773979187},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.45730000734329224},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4562000036239624},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4440999925136566},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.4318000078201294},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4106999933719635},{"id":"https://openalex.org/C4199805","wikidata":"https://www.wikidata.org/wiki/Q2725903","display_name":"Gaussian noise","level":2,"score":0.39340001344680786},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.3864000141620636},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.361299991607666},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.32989999651908875},{"id":"https://openalex.org/C148043351","wikidata":"https://www.wikidata.org/wiki/Q4456944","display_name":"Current (fluid)","level":2,"score":0.32659998536109924},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3125},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.2840000092983246},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.2809999883174896},{"id":"https://openalex.org/C29265498","wikidata":"https://www.wikidata.org/wiki/Q7047719","display_name":"Noise measurement","level":3,"score":0.2770000100135803},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2705000042915344},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2529999911785126},{"id":"https://openalex.org/C65660741","wikidata":"https://www.wikidata.org/wiki/Q3952743","display_name":"Score","level":2,"score":0.2524999976158142}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.23171","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23171","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.23171","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23171","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.8243902921676636}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recent":[0],"advancements":[1],"in":[2,72,176],"instructional":[3],"fine-tuning":[4,64,98,178],"have":[5],"injected":[6],"noise":[7,27,57,71],"into":[8,170],"embeddings,":[9],"with":[10],"NEFTune":[11,135],"(Jain":[12],"et":[13],"al.,":[14],"2024)":[15],"setting":[16],"benchmarks":[17],"using":[18,102,122],"uniform":[19,26],"noise.":[20],"Despite":[21],"NEFTune's":[22],"empirical":[23],"findings":[24],"that":[25],"outperforms":[28,155],"Gaussian":[29],"noise,":[30],"the":[31,79,93,99,132,164,171,177,196],"reasons":[32],"for":[33,66],"this":[34,42,117,190],"remain":[35],"unclear.":[36],"This":[37,74,126],"paper":[38],"aims":[39,76],"to":[40,77,120],"clarify":[41],"by":[43,82],"offering":[44],"a":[45,62,107,128],"thorough":[46],"analysis,":[47],"both":[48],"theoretical":[49],"and":[50,143],"empirical,":[51],"indicating":[52],"comparable":[53],"performance":[54,91],"among":[55],"these":[56],"types.":[58],"Additionally,":[59],"we":[60],"introduce":[61],"new":[63],"method":[65,75],"language":[67,180],"models,":[68],"utilizing":[69],"symmetric":[70,123],"embeddings.":[73,125],"enhance":[78],"model's":[80],"function":[81],"more":[83,167],"stringently":[84],"regulating":[85],"its":[86],"local":[87],"curvature,":[88],"demonstrating":[89],"superior":[90],"over":[92,131,195],"current":[94,158],"method,":[95,134],"NEFTune.":[96,156],"When":[97],"LLaMA-2-7B":[100],"model":[101],"Alpaca,":[103],"standard":[104],"techniques":[105],"yield":[106],"29.79%":[108],"score":[109,118],"on":[110,140],"AlpacaEval.":[111],"However,":[112],"our":[113],"approach,":[114,183],"SymNoise,":[115,184],"increases":[116],"significantly":[119],"69.04%,":[121],"noisy":[124],"is":[127,185],"6.7%":[129],"improvement":[130,194],"state-of-the-art":[133,198],"(64.69%).":[136],"Furthermore,":[137],"when":[138],"tested":[139],"various":[141],"models":[142],"stronger":[144],"baseline":[145],"instruction":[146],"datasets,":[147],"such":[148],"as":[149],"Evol-Instruct,":[150],"ShareGPT,":[151],"OpenPlatypus,":[152],"SymNoise":[153],"consistently":[154],"The":[157],"literature,":[159],"including":[160],"NEFTune,":[161],"has":[162],"underscored":[163],"importance":[165],"of":[166,173,179],"in-depth":[168],"research":[169],"application":[172],"noise-based":[174],"strategies":[175],"models.":[181],"Our":[182],"another":[186],"significant":[187],"step":[188],"towards":[189],"direction,":[191],"showing":[192],"notable":[193],"existing":[197],"method.":[199]},"counts_by_year":[],"updated_date":"2026-05-26T06:23:06.488380","created_date":"2026-05-26T00:00:00"}