{"id":"https://openalex.org/W7161000475","doi":"https://doi.org/10.48550/arxiv.2605.11936","title":"From Noise to Diversity: Random Embedding Injection in LLM Reasoning","display_name":"From Noise to Diversity: Random Embedding Injection in LLM Reasoning","publication_year":2026,"publication_date":"2026-05-12","ids":{"openalex":"https://openalex.org/W7161000475","doi":"https://doi.org/10.48550/arxiv.2605.11936"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.11936","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11936","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":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.2605.11936","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136085965","display_name":"Heejun Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Heejun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114137721","display_name":"Seungpil Lee","orcid":"https://orcid.org/0009-0007-1133-0848"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Seungpil","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122947792","display_name":"Jewon Yeom","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yeom, Jewon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122972619","display_name":"Jaewon Sok","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sok, Jaewon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136014706","display_name":"Seonghyeon Park","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Park, Seonghyeon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122968342","display_name":"Jeongjae Park","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Park, Jeongjae","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136015345","display_name":"Taesup Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Taesup","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5073596347","display_name":"Sundong Kim","orcid":"https://orcid.org/0000-0001-9687-2409"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Sundong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":8,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2126999944448471,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2126999944448471,"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/T10028","display_name":"Topic Modeling","score":0.10689999908208847,"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/T11596","display_name":"Constraint Satisfaction and Optimization","score":0.07180000096559525,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/embedding","display_name":"Embedding","score":0.7682999968528748},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4722999930381775},{"id":"https://openalex.org/keywords/lift","display_name":"Lift (data mining)","score":0.46700000762939453},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.4596000015735626},{"id":"https://openalex.org/keywords/variance","display_name":"Variance (accounting)","score":0.426800012588501},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.4115999937057495},{"id":"https://openalex.org/keywords/gaussian","display_name":"Gaussian","score":0.4088999927043915},{"id":"https://openalex.org/keywords/random-sequence","display_name":"Random sequence","score":0.40049999952316284},{"id":"https://openalex.org/keywords/fidelity","display_name":"Fidelity","score":0.397599995136261},{"id":"https://openalex.org/keywords/probability-distribution","display_name":"Probability distribution","score":0.366100013256073}],"concepts":[{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.7682999968528748},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5536999702453613},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5327000021934509},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4722999930381775},{"id":"https://openalex.org/C139002025","wikidata":"https://www.wikidata.org/wiki/Q3001212","display_name":"Lift (data mining)","level":2,"score":0.46700000762939453},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.4596000015735626},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.426800012588501},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.4115999937057495},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.4088999927043915},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.40790000557899475},{"id":"https://openalex.org/C186515451","wikidata":"https://www.wikidata.org/wiki/Q228191","display_name":"Random sequence","level":3,"score":0.40049999952316284},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.397599995136261},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.366100013256073},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.36340001225471497},{"id":"https://openalex.org/C184050105","wikidata":"https://www.wikidata.org/wiki/Q273163","display_name":"Isotropy","level":2,"score":0.33059999346733093},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3269999921321869},{"id":"https://openalex.org/C2776240099","wikidata":"https://www.wikidata.org/wiki/Q327018","display_name":"Interrogation","level":2,"score":0.3237999975681305},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.32109999656677246},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.3172999918460846},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3109999895095825},{"id":"https://openalex.org/C61224824","wikidata":"https://www.wikidata.org/wiki/Q2260434","display_name":"Mixture model","level":2,"score":0.3084999918937683},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3005000054836273},{"id":"https://openalex.org/C57098296","wikidata":"https://www.wikidata.org/wiki/Q12021746","display_name":"Interrogative","level":2,"score":0.29910001158714294},{"id":"https://openalex.org/C2778029271","wikidata":"https://www.wikidata.org/wiki/Q5421931","display_name":"Extension (predicate logic)","level":2,"score":0.2976999878883362},{"id":"https://openalex.org/C125112378","wikidata":"https://www.wikidata.org/wiki/Q176640","display_name":"Randomness","level":2,"score":0.29249998927116394},{"id":"https://openalex.org/C2986577269","wikidata":"https://www.wikidata.org/wiki/Q11306265","display_name":"Random noise","level":2,"score":0.2799000144004822},{"id":"https://openalex.org/C13655849","wikidata":"https://www.wikidata.org/wiki/Q25294","display_name":"Hammer","level":2,"score":0.275299996137619},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.2750999927520752},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.27140000462532043},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2709999978542328},{"id":"https://openalex.org/C9376300","wikidata":"https://www.wikidata.org/wiki/Q168817","display_name":"Algebraic number","level":2,"score":0.2526000142097473},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.25110000371932983}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.11936","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11936","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":"doi:10.48550/arxiv.2605.11936","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.11936","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":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[{"display_name":"Quality Education","score":0.6838647723197937,"id":"https://metadata.un.org/sdg/4"}],"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],"soft":[1,96,202],"prompt":[2,203],"research":[3],"has":[4,32,113],"tried":[5],"to":[6,59,72,94,114,145,245],"improve":[7],"reasoning":[8,100,130],"by":[9],"inserting":[10],"trained":[11],"vectors":[12,58],"into":[13,185],"LLM":[14],"inputs,":[15],"yet":[16,90],"whether":[17],"the":[18,22,27,45,60,73,79,83,120,123,142,166,182,198,214,236],"gain":[19],"comes":[20],"from":[21,26,67,243],"learned":[23,87],"content":[24],"or":[25],"act":[28],"of":[29,55,78,173,201,217,235],"injection":[30,218],"itself":[31],"not":[33],"been":[34],"carefully":[35],"separated.":[36],"We":[37,149],"study":[38],"Random":[39],"Soft":[40],"Prompts":[41],"(RSPs),":[42],"which":[43],"drop":[44],"training":[46,187,224],"step":[47],"entirely":[48],"and":[49,76,89,129,133,188,225,232,239],"append":[50],"a":[51,116,146,210,230],"freshly":[52,206],"drawn":[53],"sequence":[54,84],"random":[56,118],"embedding":[57,81],"input.":[61],"Each":[62],"RSP":[63,196],"vector":[64],"is":[65,176],"sampled":[66],"an":[68,241],"isotropic":[69],"Gaussian":[70],"fitted":[71],"entrywise":[74],"mean":[75],"variance":[77],"pretrained":[80],"table;":[82],"carries":[85],"no":[86],"content,":[88],"reaches":[91],"accuracy":[92],"comparable":[93],"optimized":[95],"prompts":[97],"on":[98],"math":[99],"benchmarks":[101],"in":[102,108,223],"several":[103],"settings.":[104],"The":[105],"mechanism":[106],"unfolds":[107],"two":[109],"stages:":[110],"because":[111],"attention":[112],"absorb":[115],"never-seen-before":[117],"position,":[119],"distribution":[121],"over":[122],"first":[124],"few":[125],"generated":[126],"tokens":[127],"flattens":[128],"trajectories":[131],"branch,":[132],"as":[134],"generation":[135],"continues":[136],"this":[137],"influence":[138],"dilutes":[139],"naturally":[140],"so":[141],"response":[143],"commits":[144],"single":[147],"completion.":[148],"show":[150],"that":[151,168,219],"during":[152],"inference":[153,244],"RSPs":[154],"lift":[155],"early-stage":[156],"token":[157],"diversity":[158],"and,":[159],"combined":[160],"with":[161],"temperature":[162],"sampling,":[163],"widen":[164],"Pass@N,":[165],"probability":[167],"at":[169],"least":[170],"one":[171],"out":[172],"N":[174],"attempts":[175],"correct.":[177],"Beyond":[178],"inference,":[179],"we":[180],"carry":[181],"same":[183],"effect":[184,216],"DAPO":[186],"demonstrate":[189],"practical":[190],"gains.":[191],"Our":[192],"contributions":[193],"are:":[194],"(i)":[195],"isolates":[197],"simplest":[199],"form":[200,226],"--":[204,208],"training-free,":[205],"resampled":[207],"providing":[209],"unified":[211],"lens":[212],"for":[213],"structural":[215],"variants":[220],"otherwise":[221],"differing":[222],"all":[227],"share;":[228],"(ii)":[229],"theoretical":[231],"empirical":[233],"validation":[234],"underlying":[237],"mechanism;":[238],"(iii)":[240],"extension":[242],"training.":[246]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-14T00:00:00"}