{"id":"https://openalex.org/W7127080975","doi":"https://doi.org/10.48550/arxiv.2601.22805","title":"SOMBRERO: Measuring and Steering Boundary Placement in End-to-End Hierarchical Sequence Models","display_name":"SOMBRERO: Measuring and Steering Boundary Placement in End-to-End Hierarchical Sequence Models","publication_year":2026,"publication_date":"2026-01-30","ids":{"openalex":"https://openalex.org/W7127080975","doi":"https://doi.org/10.48550/arxiv.2601.22805"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2601.22805","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.22805","pdf_url":"https://arxiv.org/pdf/2601.22805","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2601.22805","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5115944809","display_name":"Pit Neitemeier","orcid":"https://orcid.org/0000-0003-0438-7831","observed_orcids":["https://orcid.org/0000-0003-0438-7831"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Neitemeier, Pit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124857076","display_name":"Alessio Serra","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Serra, Alessio","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124785502","display_name":"Jiaze Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Jiaze","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070078683","display_name":"Sascha Wirges","orcid":"https://orcid.org/0009-0003-3316-4140","observed_orcids":["https://orcid.org/0009-0003-3316-4140"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wirges, Sascha","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047338615","display_name":"Lukas Balles","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Balles, Lukas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5049054914","display_name":"Jan Hendrik Metzen","orcid":"https://orcid.org/0000-0002-6349-7806","observed_orcids":["https://orcid.org/0000-0002-6349-7806"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Metzen, Jan Hendrik","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"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/T10028","display_name":"Topic Modeling","score":0.8593999743461609,"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":0.8593999743461609,"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":0.1136000007390976,"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/T11269","display_name":"Algorithms and Data Compression","score":0.004999999888241291,"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/computational-efficiency","display_name":"computational efficiency","score":0.49900001287460327},{"id":"https://openalex.org/keywords/tokenization","display_name":"tokenization","score":0.42800000309944153},{"id":"https://openalex.org/keywords/sombrero","display_name":"sombrero","score":0.6399999856948853}],"concepts":[{"id":"https://openalex.org/C62354387","wikidata":"https://www.wikidata.org/wiki/Q875399","display_name":"Boundary (topology)","level":2,"score":0.744700014591217},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6833000183105469},{"id":"https://openalex.org/C3770464","wikidata":"https://www.wikidata.org/wiki/Q775963","display_name":"Smoothing","level":2,"score":0.6194000244140625},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.6158999800682068},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.579200029373169},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.5590000152587891},{"id":"https://openalex.org/C43364308","wikidata":"https://www.wikidata.org/wiki/Q8799","display_name":"Byte","level":2,"score":0.4449000060558319},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4429999887943268},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.43389999866485596},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41690000891685486},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3889000117778778},{"id":"https://openalex.org/C182310444","wikidata":"https://www.wikidata.org/wiki/Q1332643","display_name":"Boundary value problem","level":2,"score":0.35280001163482666},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3400999903678894},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.33869999647140503},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.3059999942779541},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.30570000410079956},{"id":"https://openalex.org/C36299963","wikidata":"https://www.wikidata.org/wiki/Q1369844","display_name":"Observability","level":2,"score":0.29580000042915344},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2720000147819519},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2669000029563904},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.257099986076355}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2601.22805","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.22805","pdf_url":"https://arxiv.org/pdf/2601.22805","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.2601.22805","doi":"https://doi.org/10.48550/arxiv.2601.22805","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2601.22805","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:2601.22805","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.22805","pdf_url":"https://arxiv.org/pdf/2601.22805","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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4414942650","https://openalex.org/W7221082090","https://openalex.org/W7131798792","https://openalex.org/W3172862365","https://openalex.org/W1608076529","https://openalex.org/W7096807775","https://openalex.org/W2420187884","https://openalex.org/W172682952","https://openalex.org/W4389520445","https://openalex.org/W4385573212"],"abstract_inverted_index":{"Hierarchical":[0],"sequence":[1],"models":[2],"replace":[3],"fixed":[4],"tokenization":[5],"with":[6,64,135],"learned":[7],"segmentations":[8],"that":[9,130],"compress":[10],"long":[11],"byte":[12],"sequences":[13],"for":[14],"efficient":[15],"autoregressive":[16],"modeling.":[17],"While":[18],"recent":[19],"end-to-end":[20],"methods":[21],"can":[22],"learn":[23],"meaningful":[24],"boundaries":[25,129],"from":[26],"the":[27,96,124],"language-modeling":[28],"objective":[29],"alone,":[30],"it":[31],"remains":[32],"difficult":[33],"to":[34],"quantitatively":[35],"assess":[36],"and":[37,87,112,119,127],"systematically":[38],"steer":[39],"where":[40],"compute":[41,134],"is":[42],"spent.":[43],"We":[44],"introduce":[45],"a":[46,83],"router-agnostic":[47],"metric":[48],"of":[49],"boundary":[50,52,77,85,89],"quality,":[51],"enrichment":[53],"B,":[54],"which":[55,75],"measures":[56],"how":[57],"strongly":[58],"chunk":[59],"starts":[60],"concentrate":[61],"on":[62,101],"positions":[63],"high":[65],"next-byte":[66],"surprisal.":[67],"Guided":[68],"by":[69,91],"this":[70],"metric,":[71],"we":[72],"propose":[73],"Sombrero,":[74],"steers":[76],"placement":[78],"toward":[79],"predictive":[80],"difficulty":[81],"via":[82],"confidence-alignment":[84],"loss":[86],"stabilizes":[88],"learning":[90],"applying":[92],"confidence-weighted":[93],"smoothing":[94],"at":[95],"input":[97],"level":[98],"rather":[99],"than":[100],"realized":[102],"chunks.":[103],"On":[104],"1B":[105],"scale,":[106],"across":[107],"UTF-8":[108],"corpora":[109],"covering":[110],"English":[111],"German":[113],"text":[114],"as":[115,117],"well":[116],"code":[118],"mathematical":[120],"content,":[121],"Sombrero":[122],"improves":[123],"accuracy-efficiency":[125],"trade-off":[126],"yields":[128],"more":[131],"consistently":[132],"align":[133],"hard-to-predict":[136],"positions.":[137]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-03T00:00:00"}