{"id":"https://openalex.org/W7129019840","doi":"https://doi.org/10.48550/arxiv.2602.12403","title":"MonoLoss: A Training Objective for Interpretable Monosemantic Representations","display_name":"MonoLoss: A Training Objective for Interpretable Monosemantic Representations","publication_year":2026,"publication_date":"2026-02-12","ids":{"openalex":"https://openalex.org/W7129019840","doi":"https://doi.org/10.48550/arxiv.2602.12403"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.12403","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5095848454","display_name":"Ali Nasiri-Sarvi","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Nasiri-Sarvi, Ali","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Nguyen, Anh Tien","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nguyen, Anh Tien","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077743201","display_name":"Hassan Rivaz","orcid":"https://orcid.org/0000-0001-5800-3034"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rivaz, Hassan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126122455","display_name":"Dimitris Samaras","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Samaras, Dimitris","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5126152515","display_name":"Mahdi S. Hosseini","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hosseini, Mahdi S.","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5095848454"],"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.34279999136924744,"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"}},"topics":[{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.34279999136924744,"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"}},{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.17960000038146973,"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":0.11829999834299088,"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/speedup","display_name":"Speedup","score":0.8155999779701233},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.7250000238418579},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.7070000171661377},{"id":"https://openalex.org/keywords/pairwise-comparison","display_name":"Pairwise comparison","score":0.7049999833106995},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.5795999765396118},{"id":"https://openalex.org/keywords/class","display_name":"Class (philosophy)","score":0.5490999817848206},{"id":"https://openalex.org/keywords/fraction","display_name":"Fraction (chemistry)","score":0.5414999723434448},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.499099999666214},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.46459999680519104}],"concepts":[{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.8155999779701233},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.7250000238418579},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.7070000171661377},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.7049999833106995},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6700999736785889},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.628600001335144},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.6182000041007996},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.5795999765396118},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.5490999817848206},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.5414999723434448},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.499099999666214},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.46459999680519104},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.4569999873638153},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.38519999384880066},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3734999895095825},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.37220001220703125},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.36340001225471497},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.36169999837875366},{"id":"https://openalex.org/C101738243","wikidata":"https://www.wikidata.org/wiki/Q786435","display_name":"Autoencoder","level":3,"score":0.3366999924182892},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.2985999882221222},{"id":"https://openalex.org/C45235069","wikidata":"https://www.wikidata.org/wiki/Q278425","display_name":"Table (database)","level":2,"score":0.2921999990940094},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.27250000834465027},{"id":"https://openalex.org/C2776299755","wikidata":"https://www.wikidata.org/wiki/Q432449","display_name":"Carry (investment)","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.26919999718666077},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C2780589192","wikidata":"https://www.wikidata.org/wiki/Q7285140","display_name":"Raising (metalworking)","level":2,"score":0.25099998712539673}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.12403","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},{"id":"doi:10.48550/arxiv.2602.12403","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.12403","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.12403","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Sparse":[0],"autoencoders":[1],"(SAEs)":[2],"decompose":[3],"polysemantic":[4],"neural":[5],"representations,":[6],"where":[7],"neurons":[8],"respond":[9],"to":[10,87,108,171,190,205],"multiple":[11],"unrelated":[12],"concepts,":[13],"into":[14],"monosemantic":[15,133,212],"features":[16],"that":[17,60,70,124],"capture":[18],"single,":[19],"interpretable":[20,132],"concepts.":[21],"However,":[22],"standard":[23,216],"training":[24,46,113],"objectives":[25],"only":[26,101],"weakly":[27],"encourage":[28],"this":[29],"decomposition,":[30],"and":[31,47,55,95,141,148,178,199,211],"existing":[32],"monosemanticity":[33],"metrics":[34],"require":[35],"pairwise":[36],"comparisons":[37],"across":[38,175],"all":[39,176],"dataset":[40,80],"samples,":[41],"making":[42],"them":[43],"inefficient":[44],"during":[45,97,197],"evaluation.":[48],"We":[49],"study":[50],"a":[51,57,68,88,112,121,166],"recent":[52],"MonoScore":[53,110,153],"metric":[54],"derive":[56],"single-pass":[58],"algorithm":[59],"computes":[61],"exactly":[62],"the":[63,77,117,182],"same":[64],"quantity,":[65],"but":[66],"with":[67,76,181],"cost":[69],"grows":[71],"linearly,":[72],"rather":[73],"than":[74],"quadratically,":[75],"number":[78],"of":[79,165],"images.":[81],"On":[82],"OpenImagesV7,":[83],"we":[84,115],"achieve":[85],"up":[86,204],"1200x":[89],"speedup":[90,92],"wall-clock":[91],"in":[93],"evaluation":[94],"159x":[96],"training,":[98],"while":[99],"adding":[100],"~4%":[102],"per-epoch":[103],"overhead.":[104],"This":[105],"allows":[106],"us":[107],"treat":[109],"as":[111,193],"signal:":[114],"introduce":[116],"Monosemanticity":[118],"Loss":[119],"(MonoLoss),":[120],"plug-in":[122],"objective":[123],"directly":[125],"rewards":[126],"semantically":[127],"consistent":[128],"activations":[129],"for":[130,154],"learning":[131],"representations.":[134],"Across":[135],"SAEs":[136],"trained":[137],"on":[138,209,215],"CLIP,":[139],"SigLIP2,":[140],"pretrained":[142],"ViT":[143],"features,":[144],"using":[145],"BatchTopK,":[146],"TopK,":[147],"JumpReLU":[149],"SAEs,":[150],"MonoLoss":[151,157,202],"increases":[152],"most":[155],"latents.":[156],"also":[158],"consistently":[159],"improves":[160],"class":[161],"purity":[162,187],"(the":[163],"fraction":[164],"latent's":[167],"activating":[168,213],"images":[169],"belonging":[170],"its":[172],"dominant":[173],"class)":[174],"encoder":[177],"SAE":[179],"combinations,":[180],"largest":[183],"gain":[184],"raising":[185],"baseline":[186],"from":[188],"0.152":[189],"0.723.":[191],"Used":[192],"an":[194],"auxiliary":[195],"regularizer":[196],"ResNet-50":[198],"CLIP-ViT-B/32":[200],"finetuning,":[201],"yields":[203],"0.6\\%":[206],"accuracy":[207],"gains":[208],"ImageNet-1K":[210],"patterns":[214],"benchmark":[217],"datasets.":[218],"The":[219],"code":[220],"is":[221],"publicly":[222],"available":[223],"at":[224],"https://github.com/AtlasAnalyticsLab/MonoLoss.":[225]},"counts_by_year":[],"updated_date":"2026-04-23T09:07:50.710637","created_date":"2026-02-17T00:00:00"}