{"id":"https://openalex.org/W7213358501","doi":"https://doi.org/10.48550/arxiv.2609.14127","title":"Exact Finite Attention Responses From RoPE Derivatives","display_name":"Exact Finite Attention Responses From RoPE Derivatives","publication_year":2026,"publication_date":"2026-09-12","ids":{"openalex":"https://openalex.org/W7213358501","doi":"https://doi.org/10.48550/arxiv.2609.14127"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.14127","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.14127","pdf_url":"https://arxiv.org/pdf/2609.14127","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.14127","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5150887435","display_name":"Julie Huang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Julie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119958722","display_name":"Maggie Chlon","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chlon, Maggie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129735299","display_name":"Gregory Gutin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I124357947","display_name":"University of London","ror":"https://ror.org/04cw6st05","country_code":"GB","type":"education","lineage":["https://openalex.org/I124357947"]},{"id":"https://openalex.org/I184558857","display_name":"Royal Holloway University of London","ror":"https://ror.org/04g2vpn86","country_code":"GB","type":"education","lineage":["https://openalex.org/I124357947","https://openalex.org/I184558857"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Gutin, Gregory","raw_affiliation_strings":["Department of Computer Science , Royal Holloway , University of London"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science , Royal Holloway , University of London","institution_ids":["https://openalex.org/I124357947","https://openalex.org/I184558857"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5119958721","display_name":"Leon Chlon","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I40120149","display_name":"University of Oxford","ror":"https://ror.org/052gg0110","country_code":"GB","type":"education","lineage":["https://openalex.org/I40120149"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Chlon, Leon","raw_affiliation_strings":["University of Oxford"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Oxford","institution_ids":["https://openalex.org/I40120149"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"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":0.9740999937057495,"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":0.9740999937057495,"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":0.01640000008046627,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.0024999999441206455,"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/qwen","display_name":"Qwen","score":0.5099999904632568}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5216000080108643},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.4821999967098236},{"id":"https://openalex.org/C139676723","wikidata":"https://www.wikidata.org/wiki/Q1193832","display_name":"Sign (mathematics)","level":2,"score":0.44609999656677246},{"id":"https://openalex.org/C159718280","wikidata":"https://www.wikidata.org/wiki/Q5526353","display_name":"Gas meter prover","level":3,"score":0.4390000104904175},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4235000014305115},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.42260000109672546},{"id":"https://openalex.org/C129844170","wikidata":"https://www.wikidata.org/wiki/Q41299","display_name":"Quadratic equation","level":2,"score":0.4131999909877777},{"id":"https://openalex.org/C200331156","wikidata":"https://www.wikidata.org/wiki/Q506041","display_name":"Jacobian matrix and determinant","level":2,"score":0.4041000008583069},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.397599995136261},{"id":"https://openalex.org/C96865113","wikidata":"https://www.wikidata.org/wiki/Q2946816","display_name":"Certificate","level":2,"score":0.391400009393692},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.38850000500679016},{"id":"https://openalex.org/C126780896","wikidata":"https://www.wikidata.org/wiki/Q899871","display_name":"Distortion (music)","level":4,"score":0.3797999918460846},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.366100013256073},{"id":"https://openalex.org/C2780012671","wikidata":"https://www.wikidata.org/wiki/Q1343870","display_name":"Corollary","level":2,"score":0.3537999987602234},{"id":"https://openalex.org/C74050887","wikidata":"https://www.wikidata.org/wiki/Q848368","display_name":"Rotation (mathematics)","level":2,"score":0.34689998626708984},{"id":"https://openalex.org/C138187205","wikidata":"https://www.wikidata.org/wiki/Q131251","display_name":"Tangent","level":2,"score":0.3366999924182892},{"id":"https://openalex.org/C181330731","wikidata":"https://www.wikidata.org/wiki/Q2068418","display_name":"Finite difference","level":2,"score":0.33230000734329224},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.3314000070095062},{"id":"https://openalex.org/C195065555","wikidata":"https://www.wikidata.org/wiki/Q214881","display_name":"Curvature","level":2,"score":0.30550000071525574},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.3034999966621399},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.2978000044822693},{"id":"https://openalex.org/C149364088","wikidata":"https://www.wikidata.org/wiki/Q185917","display_name":"Translation (biology)","level":4,"score":0.2976999878883362},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.2879999876022339},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.2824999988079071},{"id":"https://openalex.org/C61224824","wikidata":"https://www.wikidata.org/wiki/Q2260434","display_name":"Mixture model","level":2,"score":0.27480000257492065},{"id":"https://openalex.org/C154690210","wikidata":"https://www.wikidata.org/wiki/Q1668499","display_name":"Rewriting","level":2,"score":0.27239999175071716},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.2660999894142151},{"id":"https://openalex.org/C2781300812","wikidata":"https://www.wikidata.org/wiki/Q178061","display_name":"Shock (circulatory)","level":2,"score":0.2660999894142151},{"id":"https://openalex.org/C205203396","wikidata":"https://www.wikidata.org/wiki/Q612143","display_name":"Bilinear interpolation","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.2596000134944916},{"id":"https://openalex.org/C32900221","wikidata":"https://www.wikidata.org/wiki/Q181365","display_name":"Dot product","level":2,"score":0.25929999351501465},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C47890412","wikidata":"https://www.wikidata.org/wiki/Q1179296","display_name":"Tangent vector","level":3,"score":0.2556999921798706},{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.2517000138759613}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.14127","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.14127","pdf_url":"https://arxiv.org/pdf/2609.14127","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.14127","doi":"https://doi.org/10.48550/arxiv.2609.14127","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.14127","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.14127","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.14127","pdf_url":"https://arxiv.org/pdf/2609.14127","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7213358501.pdf","grobid_xml":"https://content.openalex.org/works/W7213358501.grobid-xml"},"referenced_works_count":8,"referenced_works":["https://openalex.org/W1971713783","https://openalex.org/W4211077197","https://openalex.org/W4385571886","https://openalex.org/W4388481390","https://openalex.org/W4388979610","https://openalex.org/W4392428679","https://openalex.org/W4404783717","https://openalex.org/W4415798413"],"related_works":["https://openalex.org/W7211995705","https://openalex.org/W7092213973","https://openalex.org/W7222879169","https://openalex.org/W6893319214","https://openalex.org/W7172522578","https://openalex.org/W4411119665","https://openalex.org/W7213525991","https://openalex.org/W7163643990","https://openalex.org/W4405767785","https://openalex.org/W7170809026"],"abstract_inverted_index":{"We":[0],"derive":[1],"exact":[2,181],"local":[3,177,205],"responses":[4],"for":[5,176],"attention":[6,152],"interventions,":[7],"allowing":[8],"candidate":[9,198],"edits":[10,64],"to":[11],"be":[12],"scored":[13],"from":[14],"a":[15,120,128,142,160,173],"cached":[16],"baseline":[17,161],"and":[18,79,87,137,163,191,200],"one":[19],"backward":[20],"pass.":[21],"The":[22,53,169],"starting":[23],"point":[24],"is":[25,105],"the":[26,37,45,76,90,95,150],"RoPE":[27],"derivative":[28],"$\\partial_p":[29],"z(p)":[30],"=":[31,99],"A":[32],"z(p)$:":[33],"its":[34],"integral":[35],"gives":[36],"finite":[38],"positional":[39,63,77],"displacement,":[40],"which":[41,104],"we":[42],"carry":[43],"through":[44],"softmax":[46],"without":[47],"linearising":[48],"either":[49],"rotation":[50],"or":[51],"normalisation.":[52],"resulting":[54],"predictions":[55,155],"achieve":[56],"95.36--96.52%":[57],"sign":[58],"accuracy":[59],"across":[60,131],"92,160":[61],"executed":[62],"on":[65],"768":[66],"held-out":[67],"prompt":[68],"sets,":[69],"reducing":[70],"answer-margin":[71],"MAE":[72,116],"by":[73,80,107,117,166],"73.6--82.5%":[74],"against":[75,82,141],"Jacobian":[78],"36.2--50.9%":[81],"zero.":[83],"For":[84],"simultaneous":[85],"key":[86],"value":[88],"edits,":[89],"same":[91],"divided-difference":[92],"calculus":[93,170],"isolates":[94],"interaction":[96,144],"term":[97],"$C_{KV}":[98],"\\sum_j":[100],"(p'_j":[101],"-":[102],"p_j)\\,\\varepsilon_j$,":[103],"omitted":[106],"adding":[108],"separate":[109],"attributions.":[110],"Retaining":[111],"it":[112],"reduces":[113],"downstream":[114,154],"margin":[115],"more":[118],"than":[119],"factor":[121],"of":[122,127],"nine":[123],"in":[124],"every":[125],"setting":[126],"5,120-intervention":[129],"sweep":[130],"two":[132,135],"Qwen":[133],"sizes,":[134],"tasks,":[136],"multiple":[138],"layers;":[139],"reductions":[140],"quadratic":[143],"correction":[145],"are":[146,164],"75.9--98.5%.":[147],"Exactness":[148],"concerns":[149],"edited":[151],"write;":[153],"contract":[156],"that":[157],"response":[158],"with":[159],"gradient":[162],"evaluated":[165],"native":[167],"execution.":[168],"also":[171],"yields":[172],"KL":[174],"certificate":[175],"approximation":[178],"error,":[179],"an":[180],"query-conditioned":[182],"gradient-step":[183],"representation":[184],"whose":[185],"curvature":[186],"identifies":[187],"attention-preserving":[188],"query":[189,193],"directions,":[190],"minimum-norm":[192],"control.":[194],"Sparse":[195],"evaluation":[196],"supports":[197],"ranking":[199],"cache":[201],"decisions":[202],"under":[203],"explicit":[204],"distortion":[206],"criteria.":[207]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-16T00:00:00"}