{"id":"https://openalex.org/W7225416154","doi":"https://doi.org/10.5281/zenodo.24014129","title":"Language as a Compression Artifact of Physical Reality: Category-Theoretic Nuisance Fibrations, the Platonic Representation Theorem, Procrustes Spectral Decay, and Cross-Scale Activation Patching","display_name":"Language as a Compression Artifact of Physical Reality: Category-Theoretic Nuisance Fibrations, the Platonic Representation Theorem, Procrustes Spectral Decay, and Cross-Scale Activation Patching","publication_year":2026,"publication_date":"2026-11-06","ids":{"openalex":"https://openalex.org/W7225416154","doi":"https://doi.org/10.5281/zenodo.24014129"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.24014129","doi":"https://doi.org/10.5281/zenodo.24014129","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24014129","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.24014129","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139077210","display_name":"Jeremiah Swann","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Swann, Jeremiah","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5139077210"],"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":0.5526999831199646,"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":0.5526999831199646,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.08049999922513962,"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":0.05310000106692314,"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/markov-categories","display_name":"Markov categories","score":0.7440000176429749},{"id":"https://openalex.org/keywords/jepa","display_name":"JEPA","score":0.6069999933242798},{"id":"https://openalex.org/keywords/information-bottleneck","display_name":"information bottleneck","score":0.7760000228881836},{"id":"https://openalex.org/keywords/next-token-prediction","display_name":"next-token prediction","score":0.3930000066757202}],"concepts":[{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5408999919891357},{"id":"https://openalex.org/C2777984123","wikidata":"https://www.wikidata.org/wiki/Q9248237","display_name":"Pointwise","level":2,"score":0.4726000130176544},{"id":"https://openalex.org/C60008888","wikidata":"https://www.wikidata.org/wiki/Q6031013","display_name":"Information bottleneck method","level":3,"score":0.4569999873638153},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4278999865055084},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.41769999265670776},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.41510000824928284},{"id":"https://openalex.org/C22789450","wikidata":"https://www.wikidata.org/wiki/Q420904","display_name":"Singular value decomposition","level":2,"score":0.38580000400543213},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.3684999942779541},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3668000102043152},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.36039999127388},{"id":"https://openalex.org/C101738243","wikidata":"https://www.wikidata.org/wiki/Q786435","display_name":"Autoencoder","level":3,"score":0.3447999954223633},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.34279999136924744},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.31520000100135803},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.30219998955726624},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2921999990940094},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.2872999906539917},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.2854999899864197},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.28299999237060547},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.27970001101493835},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.27619999647140503},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.2736999988555908},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.27250000834465027},{"id":"https://openalex.org/C47458327","wikidata":"https://www.wikidata.org/wiki/Q910404","display_name":"Random graph","level":3,"score":0.2702000141143799},{"id":"https://openalex.org/C165700671","wikidata":"https://www.wikidata.org/wiki/Q203484","display_name":"Laplace operator","level":2,"score":0.26840001344680786},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.2614000141620636},{"id":"https://openalex.org/C122280245","wikidata":"https://www.wikidata.org/wiki/Q620622","display_name":"Kernel method","level":3,"score":0.2612999975681305},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.2590999901294708},{"id":"https://openalex.org/C52622258","wikidata":"https://www.wikidata.org/wiki/Q131222","display_name":"Information theory","level":2,"score":0.2590000033378601},{"id":"https://openalex.org/C183115368","wikidata":"https://www.wikidata.org/wiki/Q856577","display_name":"Weighting","level":2,"score":0.25}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.5281/zenodo.24014129","doi":"https://doi.org/10.5281/zenodo.24014129","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24014129","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},{"id":"doi:10.5281/zenodo.24014130","doi":"https://doi.org/10.5281/zenodo.24014130","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24014130","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},{"id":"doi:10.5281/zenodo.24050391","doi":"https://doi.org/10.5281/zenodo.24050391","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24050391","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":"doi:10.5281/zenodo.24014129","doi":"https://doi.org/10.5281/zenodo.24014129","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24014129","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7140309328","https://openalex.org/W7124905663","https://openalex.org/W7165391278","https://openalex.org/W4414930467","https://openalex.org/W7128491348","https://openalex.org/W4415035641","https://openalex.org/W7201873599","https://openalex.org/W7105830470","https://openalex.org/W7139981965","https://openalex.org/W4404782820"],"abstract_inverted_index":{"Overview":[0],"This":[1],"21-page":[2,463],"research":[3],"manuscript,":[4],"theoretical":[5],"synthesis,":[6],"and":[7,20,37,106,244,277,289,339,416,444,476,488],"Python":[8,482],"benchmarking":[9,483],"suite":[10],"by":[11,179,284],"Jeremiah":[12],"Swann":[13],"(Lead":[14],"Researcher)":[15],"investigates":[16],"how":[17],"neural":[18],"language":[19,55],"vision":[21],"models":[22],"trained":[23],"via":[24,111],"gradient":[25],"descent":[26],"converge":[27],"toward":[28],"a":[29,134,210,349,357],"shared":[30],"latent":[31],"geometry\u2014and":[32],"pinpoints":[33],"the":[34,51,73,81,151,157,177,295,424,470],"exact":[35,425],"mathematical":[36],"layerwise":[38,426],"boundaries":[39],"where":[40,434],"this":[41,70],"structural":[42],"isomorphism":[43,388],"breaks":[44],"down":[45],"in":[46,50,164,167,369,389,456],"smaller-parameter":[47],"models.":[48],"Grounded":[49],"premise":[52],"that":[53,149,184,217],"human":[54],"evolved":[56],"as":[57,209],"an":[58,198,232,438],"Information":[59,91],"Bottleneck":[60,92],"(Tishby":[61],"et":[62,78,97,103,114,274,280,286,291,330,336,344,398,403,419],"al.,":[63,79,98,104,115,275,281,331,337,345,399,404,420],"2000)":[64],"compression":[65,94],"algorithm":[66],"of":[67,237,362],"physical":[68],"reality,":[69],"work":[71],"synthesizes":[72],"Platonic":[74,82,471],"Representation":[75,83,472],"Hypothesis":[76],"(Huh":[77,343],"2024),":[80],"Theorem":[84],"for":[85],"JEPAs":[86,186],"(Balestriero":[87],"&":[88,119,181,189,254,261,315,380],"LeCun,":[89,190],"2026),":[90],"semantic":[93],"dynamics":[95],"(Shani":[96],"2025),":[99,327],"Relative":[100],"Representations":[101],"(Moschella":[102],"2023),":[105],"unsupervised":[107],"cycle-consistent":[108],"embedding":[109],"translation":[110],"vec2vec":[112],"(Jha":[113],"2025).":[116,256],"Core":[117],"Theoretical":[118],"Empirical":[120],"Contributions":[121],"Category-Theoretic":[122,467],"Nuisance":[123,468],"Fibrations":[124],"($\\mathbf{1}^\\perp":[125],"=":[126,364,374,429,432],"\\mathcal{H}_S":[127],"\\oplus":[128],"\\mathcal{H}_N$):":[129],"Models":[130],"observation":[131],"modalities":[132],"inside":[133],"Markov":[135],"category":[136],"$\\mathbf{Stoch}(\\mathcal{Z},":[137],"p)$":[138],"equipped":[139],"with":[140,213,466,485],"canonical":[141],"state-projection":[142],"fibrations":[143],"$s_m":[144],":":[145,450],"\\mathcal{X}_m":[146],"\\twoheadrightarrow":[147],"\\mathcal{Z}$":[148],"separate":[150],"horizontal":[152],"world-state":[153],"subspace":[154],"$\\mathcal{H}_S$":[155],"from":[156],"vertical":[158],"modality-nuisance":[159],"fiber":[160],"$\\mathcal{H}_N$":[161],"(author":[162],"wording/syntax":[163],"text;":[165],"viewpoint/lighting":[166],"vision).":[168],"Universal":[169],"Coequalizer":[170],"Factorization":[171],"vs.":[172],"Next-Token":[173],"Bad-View":[174],"Leakage:":[175],"Integrates":[176],"proof":[178],"Balestriero":[180,188],"LeCun":[182],"(2026)":[183],"good-view":[185],"(LeJEPA,":[187],"2025)":[191],"factorize":[192],"uniquely":[193],"through":[194],"$s_m$":[195],"up":[196],"to":[197,301,422],"orthogonal":[199],"transformation":[200],"$\\mathbf{Q}_m":[201],"\\in":[202],"O(k)$,":[203],"whereas":[204],"autoregressive":[205],"next-token":[206],"prediction":[207],"acts":[208],"bad-view":[211],"JEPA":[212],"$\\mathrm{GL}(k)$":[214],"gauge":[215],"symmetry":[216],"leaks":[218],"$r":[219],"\\ge":[220,242],"1$":[221],"principal":[222],"directions":[223],"into":[224,410],"intra-document":[225],"nuisance":[226,435],"whenever":[227],"$\\beta_1":[228],">":[229],"\\alpha_k$,":[230],"imposing":[231],"irreducible":[233],"Procrustes":[234,312,323],"alignment":[235,324],"floor":[236],"$2k(1":[238],"-":[239],"\\min_m":[240],"\\rho(f_m))":[241],"2r$":[243],"requiring":[245],"$42\\times$":[246],"more":[247],"training":[248],"samples":[249],"than":[250],"good":[251],"views":[252],"(Ziyin":[253],"Chuang,":[255],"Scale":[257,297],"Splitting":[258,298],"Index":[259,299],"($\\mathsf{SSI}$)":[260],"Superposition":[262],"Disentanglement:":[263],"Extends":[264],"cross-scale":[265],"Sparse":[266],"Autoencoder":[267],"(SAE)":[268],"universality":[269],"benchmarks":[270],"across":[271,307],"Pythia":[272],"(Biderman":[273],"2023)":[276],"Gemma-2":[278],"(Riviere":[279],"2024)":[282,346,421],"established":[283],"Lan":[285],"al.":[287,292],"(2025)":[288],"Son":[290],"(2025),":[293],"defining":[294],"cumulative":[296],"($\\mathsf{SSI}_{\\text{cum}}$)":[300],"track":[302],"surjective":[303],"($\\text{Many}_F\\text{-to-}1_S$)":[304],"feature":[305],"splitting":[306],"normalized":[308],"network":[309],"depth.":[310],"Orthogonal":[311,322],"Spectral":[313],"Decay":[314],"Random":[316],"Control":[317],"Pipeline":[318],"($\\Delta_{\\text{Procrustes}}$):":[319],"Implements":[320],"closed-form":[321],"(Procrustes":[325],"Bounds,":[326],"SVCCA":[328],"(Raghu":[329],"2017),":[332],"Linear":[333],"CKA":[334],"(Kornblith":[335],"2019),":[338],"mutual":[340],"$k$-NN":[341],"$\\mathsf{CKNNA}$":[342],"paired":[347],"against":[348],"randomly":[350],"initialized":[351],"Pythia-70M":[352],"control":[353],"group":[354],"($\\mathcal{M}_{\\text{null}}$),":[355],"demonstrating":[356],"peak":[358],"learned":[359],"optimization":[360],"gain":[361],"$\\Delta_{\\text{Procrustes}}":[363],"+1.470$":[365],"($p":[366],"<":[367],"0.005$)":[368],"middle":[370],"conceptual":[371],"layers":[372],"($\\tilde{\\ell}":[373],"0.50$).":[375],"Cross-Scale":[376],"Causal":[377],"Activation":[378],"Patching":[379],"Compression":[381],"Breakdown":[382],"Boundary":[383],"($\\ell^*$):":[384],"Grounds":[385],"causal":[386,446],"functional":[387],"localized":[390],"mid-layer":[391],"MLP":[392],"associative":[393],"memory":[394],"editing":[395],"(ROME,":[396],"Meng":[397,402],"2022a;":[400],"MEMIT,":[401],"2022b),":[405],"projecting":[406],"frontier":[407],"residual":[408],"updates":[409],"smaller":[411],"experimental":[412],"cohorts":[413],"(Pythia,":[414],"Gemma-2,":[415],"Qwen2,":[417],"Yang":[418],"isolate":[423],"threshold":[427],"($\\ell^*":[428],"\\text{L5},":[430],"\\tilde{\\ell}^*":[431],"0.83$)":[433],"saturation":[436],"triggers":[437],"$\\mathsf{SSI}$":[439],"spike":[440],"($1.75":[441],"\\to":[442,452],"2.97$)":[443],"collapses":[445],"patching":[447],"efficacy":[448],"($\\eta_{\\text{patch}}":[449],"0.890":[451],"0.000$).":[453],"Included":[454],"Files":[455],"Deposit":[457],"shared_latent_space_manuscript_swann.pdf":[458],"/":[459],".tex":[460],"\u2014":[461,480,494],"Full":[462],"illustrated":[464],"manuscript":[465],"Fibrations,":[469],"Theorem,":[473],"formal":[474],"hypotheses,":[475],"empirical":[477],"telemetry.":[478],"shared_latent_space_benchmark_harness.py":[479],"Self-contained":[481],"harness":[484],"--self-test":[486],"verification":[487],"HuggingFace/PyTorch":[489],"hooks.":[490],"fig1_procrustes_random_control_delta.png,":[491],"figure_2_singular_value_decay_spectra.png,":[492],"figure_3_ssi_activation_patching_breakdown.png":[493],"300-DPI":[495],"publication":[496],"figures.":[497]},"counts_by_year":[],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2026-11-07T00:00:00"}