{"id":"https://openalex.org/W7202197839","doi":"https://doi.org/10.48550/arxiv.2608.07594","title":"Scaling Inherently Interpretable Language Models","display_name":"Scaling Inherently Interpretable Language Models","publication_year":2026,"publication_date":"2026-08-06","ids":{"openalex":"https://openalex.org/W7202197839","doi":"https://doi.org/10.48550/arxiv.2608.07594"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.07594","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.07594","pdf_url":"https://arxiv.org/pdf/2608.07594","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":"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/2608.07594","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5146479501","display_name":"Guide Labs Team","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guide Labs Team","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5146506375","display_name":"Andreas Madsen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Madsen, Andreas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5073867448","display_name":"Aya Abdelsalam Ismail","orcid":"https://orcid.org/0000-0002-3145-4385","observed_orcids":["https://orcid.org/0000-0002-3145-4385"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ismail, Aya Abdelsalam","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139917876","display_name":"Giang Nguyen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nguyen, Giang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5146493433","display_name":"Isaac Plant","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Plant, Isaac","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041654483","display_name":"Muawiz Chaudhary","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chaudhary, Muawiz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5054791538","display_name":"Nathaniel Monson","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Monson, Nathaniel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042634288","display_name":"Saqib Azim","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Azim, Saqib","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5146571298","display_name":"Zhichen Guo","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Zhichen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5112273421","display_name":"Julius Adebayo","orcid":"https://orcid.org/0009-0006-6590-1303","observed_orcids":["https://orcid.org/0009-0006-6590-1303"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Adebayo, Julius","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":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":1.0,"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":1.0,"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":0.0,"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":0.0,"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/interpretability","display_name":"interpretability","score":0.6190000176429749},{"id":"https://openalex.org/keywords/language-model-training","display_name":"language model training","score":0.6110000014305115},{"id":"https://openalex.org/keywords/diffusion-language-models","display_name":"diffusion language models","score":0.7049999833106995},{"id":"https://openalex.org/keywords/disentangled-representations","display_name":"disentangled representations","score":0.7059999704360962},{"id":"https://openalex.org/keywords/model-scaling","display_name":"model scaling","score":0.5609999895095825},{"id":"https://openalex.org/keywords/closed-loop-intervention","display_name":"closed-loop intervention","score":0.6639999747276306}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9714000225067139},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7024000287055969},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6485999822616577},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6351000070571899},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5855000019073486},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5214999914169312},{"id":"https://openalex.org/C2776036281","wikidata":"https://www.wikidata.org/wiki/Q48769818","display_name":"Constraint (computer-aided design)","level":2,"score":0.4950000047683716},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.48080000281333923},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.41819998621940613},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.35670000314712524},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.329800009727478},{"id":"https://openalex.org/C8505890","wikidata":"https://www.wikidata.org/wiki/Q605095","display_name":"Budget constraint","level":2,"score":0.2858000099658966},{"id":"https://openalex.org/C91682802","wikidata":"https://www.wikidata.org/wiki/Q620538","display_name":"Multidimensional scaling","level":2,"score":0.2840000092983246},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.27559998631477356},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.2728999853134155},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C45493050","wikidata":"https://www.wikidata.org/wiki/Q7884934","display_name":"Unified Model","level":2,"score":0.25369998812675476}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.07594","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.07594","pdf_url":"https://arxiv.org/pdf/2608.07594","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.07594","doi":"https://doi.org/10.48550/arxiv.2608.07594","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.07594","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:2608.07594","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.07594","pdf_url":"https://arxiv.org/pdf/2608.07594","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7202197839.pdf","grobid_xml":"https://content.openalex.org/works/W7202197839.grobid-xml"},"referenced_works_count":58,"referenced_works":["https://openalex.org/W2169393322","https://openalex.org/W2529194139","https://openalex.org/W2530010084","https://openalex.org/W2594475271","https://openalex.org/W2786672974","https://openalex.org/W2794325560","https://openalex.org/W2994706149","https://openalex.org/W3001279689","https://openalex.org/W3083410900","https://openalex.org/W3085380432","https://openalex.org/W3095645723","https://openalex.org/W3173476189","https://openalex.org/W3215529831","https://openalex.org/W4224210047","https://openalex.org/W4283026156","https://openalex.org/W4296932880","https://openalex.org/W4327810158","https://openalex.org/W4376167329","https://openalex.org/W4385327568","https://openalex.org/W4385681731","https://openalex.org/W4386081218","https://openalex.org/W4387225871","https://openalex.org/W4387355345","https://openalex.org/W4387725649","https://openalex.org/W4388514540","https://openalex.org/W4390041933","https://openalex.org/W4390690464","https://openalex.org/W4392822465","https://openalex.org/W4394709684","https://openalex.org/W4396821993","https://openalex.org/W4399454157","https://openalex.org/W4400517495","https://openalex.org/W4401306886","https://openalex.org/W4404313113","https://openalex.org/W4404390840","https://openalex.org/W4404404507","https://openalex.org/W4405034970","https://openalex.org/W4405655184","https://openalex.org/W4405903187","https://openalex.org/W4406032137","https://openalex.org/W4406961517","https://openalex.org/W4407309861","https://openalex.org/W4414898959","https://openalex.org/W4414946584","https://openalex.org/W4414968636","https://openalex.org/W4416091752","https://openalex.org/W4417102600","https://openalex.org/W4417297081","https://openalex.org/W4417465430","https://openalex.org/W6910504028","https://openalex.org/W6967055661","https://openalex.org/W7117161621","https://openalex.org/W7129268635","https://openalex.org/W7129372038","https://openalex.org/W7133225605","https://openalex.org/W7135156033","https://openalex.org/W7163370823","https://openalex.org/W7167078970"],"related_works":["https://openalex.org/W6928821472","https://openalex.org/W4416191647","https://openalex.org/W4415924648","https://openalex.org/W4402325998","https://openalex.org/W7213405577","https://openalex.org/W7152676055","https://openalex.org/W7134984220","https://openalex.org/W7161099702","https://openalex.org/W7128717306","https://openalex.org/W7151545533"],"abstract_inverted_index":{"Interpretability":[0],"is":[1,25],"often":[2],"treated":[3],"as":[4,13],"a":[5,39,44,98,103,166],"tax":[6],"on":[7,63,160],"capability:":[8],"language":[9,53,68,100],"models":[10,158],"are":[11],"trained":[12,159],"opaque":[14],"systems,":[15],"then":[16],"explained":[17],"after":[18],"the":[19,47,52,93,115,145],"fact,":[20],"with":[21,72,86,89,96,102,155,179],"methods":[22],"whose":[23],"reliability":[24],"difficult":[26],"to":[27,117],"establish.":[28],"In":[29],"this":[30,34],"work,":[31],"we":[32,41],"challenge":[33],"premise.":[35],"Rather":[36],"than":[37,75],"reverse-engineering":[38],"model,":[40],"make":[42],"interpretability":[43,70,170],"constraint":[45],"of":[46,59,61,110],"training":[48,124,141],"pipeline,":[49],"optimized":[50],"alongside":[51],"modeling":[54],"objective.":[55],"Across":[56],"three":[57],"orders":[58],"magnitude":[60],"compute,":[62,164],"both":[64],"autoregressive":[65],"and":[66,84,123,143,176],"diffusion":[67,99],"models,":[69],"scales":[71],"capability":[73],"rather":[74],"against":[76],"it.":[77],"Surprisingly,":[78],"model":[79,101],"representations":[80],"become":[81],"more":[82,163],"disentangled":[83],"aligned":[85],"human-understandable":[87,121],"concepts":[88],"scale.":[90,180],"We":[91],"instantiate":[92],"training-time":[94],"recipe":[95],"Steerling-8B,":[97],"causal":[104],"attention":[105],"mask.":[106],"For":[107],"any":[108],"group":[109],"generated":[111],"tokens,":[112,120],"Steerling-8B":[113,152],"attributes":[114],"output":[116,132],"relevant":[118],"input":[119],"concepts,":[122],"data.":[125],"This":[126],"enables":[127],"closed-loop":[128],"intervention:":[129],"diagnose":[130],"an":[131],"through":[133,147],"its":[134],"concept":[135,148],"or":[136],"feature":[137],"attribution,":[138],"retrieve":[139],"similar":[140],"data,":[142],"correct":[144],"behavior":[146],"steering":[149],"without":[150],"retraining.":[151],"remains":[153],"competitive":[154],"open":[156],"peer":[157],"substantially":[161],"2-16x":[162],"suggesting":[165],"different":[167],"scaling":[168],"paradigm:":[169],"can":[171],"be":[172],"designed":[173],"into":[174],"training,":[175],"it":[177],"improves":[178]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-12T00:00:00"}