{"id":"https://openalex.org/W4404334616","doi":"https://doi.org/10.48550/arxiv.2410.21272","title":"Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics","display_name":"Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics","publication_year":2024,"publication_date":"2024-10-28","ids":{"openalex":"https://openalex.org/W4404334616","doi":"https://doi.org/10.48550/arxiv.2410.21272"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2410.21272","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.21272","pdf_url":"https://arxiv.org/pdf/2410.21272","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2410.21272","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5056736847","display_name":"Yaniv Nikankin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"Nikankin, Yaniv","raw_affiliation_strings":["Technion -Israel Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion -Israel Institute of Technology","institution_ids":["https://openalex.org/I174306211"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031923942","display_name":"Anja Reusch","orcid":"https://orcid.org/0000-0002-2537-9841","observed_orcids":["https://orcid.org/0000-0002-2537-9841"]},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"Reusch, Anja","raw_affiliation_strings":["Technion -Israel Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion -Israel Institute of Technology","institution_ids":["https://openalex.org/I174306211"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020998070","display_name":"Aaron Mueller","orcid":"https://orcid.org/0009-0005-1148-5001","observed_orcids":["https://orcid.org/0009-0005-1148-5001"]},"institutions":[{"id":"https://openalex.org/I12912129","display_name":"Northeastern University","ror":"https://ror.org/04t5xt781","country_code":"US","type":"education","lineage":["https://openalex.org/I12912129"]},{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL","US"],"is_corresponding":false,"raw_author_name":"Mueller, Aaron","raw_affiliation_strings":["Northeastern University","Technion -Israel Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I12912129"]},{"raw_affiliation_string":"Technion -Israel Institute of Technology","institution_ids":["https://openalex.org/I174306211"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5051184573","display_name":"Yonatan Belinkov","orcid":"https://orcid.org/0000-0002-6280-5045","observed_orcids":["https://orcid.org/0000-0002-6280-5045"]},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"Belinkov, Yonatan","raw_affiliation_strings":["Technion -Israel Institute of Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion -Israel Institute of Technology","institution_ids":["https://openalex.org/I174306211"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":4,"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":0.8949999809265137,"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":0.8949999809265137,"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":0.045499999076128006,"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":0.005900000222027302,"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/large-language-models","display_name":"large language models","score":0.9539999961853027},{"id":"https://openalex.org/keywords/arithmetic-reasoning","display_name":"arithmetic reasoning","score":0.9879999756813049},{"id":"https://openalex.org/keywords/neuron-importance","display_name":"neuron importance","score":0.41999998688697815},{"id":"https://openalex.org/keywords/numerical-reasoning","display_name":"numerical reasoning","score":0.39899998903274536}],"concepts":[{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.8324639797210693},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5441837906837463},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.5335683226585388},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.492315411567688},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.32613667845726013},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.32021939754486084},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2410.21272","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.21272","pdf_url":"https://arxiv.org/pdf/2410.21272","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.2410.21272","doi":"https://doi.org/10.48550/arxiv.2410.21272","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2410.21272","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:2410.21272","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.21272","pdf_url":"https://arxiv.org/pdf/2410.21272","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":[{"id":"https://openalex.org/G6869880996","display_name":"Controlling Large Language Models","funder_award_id":"101165402","funder_id":"https://openalex.org/F4320338453","funder_display_name":"HORIZON EUROPE European Research Council"}],"funders":[{"id":"https://openalex.org/F4320315389","display_name":"Open Philanthropy Project","ror":"https://ror.org/004d1k391"},{"id":"https://openalex.org/F4320320300","display_name":"European Commission","ror":"https://ror.org/00k4n6c32"},{"id":"https://openalex.org/F4320323139","display_name":"Azrieli Foundation","ror":"https://ror.org/009svts64"},{"id":"https://openalex.org/F4320338453","display_name":"HORIZON EUROPE European Research Council","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4404334616.pdf","grobid_xml":"https://content.openalex.org/works/W4404334616.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7155247196","https://openalex.org/W4417104612","https://openalex.org/W7169866005","https://openalex.org/W4389519056","https://openalex.org/W4402955064","https://openalex.org/W4385572162","https://openalex.org/W4416037559","https://openalex.org/W4399837080","https://openalex.org/W7124455643","https://openalex.org/W4404088822"],"abstract_inverted_index":{"Do":[0],"large":[1],"language":[2],"models":[3],"(LLMs)":[4],"solve":[5],"reasoning":[6,26],"tasks":[7],"by":[8],"learning":[9],"robust":[10,185],"generalizable":[11],"algorithms,":[12],"or":[13],"do":[14],"they":[15,190],"memorize":[16],"training":[17],"data?":[18],"To":[19,109],"investigate":[20],"this":[21,158],"question,":[22],"we":[23,34,68,112,155],"use":[24],"arithmetic":[25,52,107,152,166,182],"as":[27,120,161],"a":[28,36,70,83,129,193],"representative":[29],"task.":[30],"Using":[31],"causal":[32],"analysis,":[33],"identify":[35],"subset":[37],"of":[38,46,64,73,96,137,147,165,195],"the":[39,47,62,94,101,134,142,148,162],"model":[40],"(a":[41],"circuit)":[42],"that":[43,76,93,122,133,144,157,179],"explains":[44,145],"most":[45,146],"model's":[48,149],"behavior":[49],"for":[50],"basic":[51],"logic":[53],"and":[54,87],"examine":[55],"its":[56],"functionality.":[57],"By":[58],"zooming":[59],"in":[60,169],"on":[61,151,192],"level":[63],"individual":[65],"circuit":[66],"neurons,":[67],"discover":[69],"sparse":[71],"set":[72],"important":[74],"neurons":[75,99,121],"implement":[77],"simple":[78],"heuristics.":[79],"Each":[80],"heuristic":[81,98,118,139],"identifies":[82],"numerical":[84],"input":[85],"pattern":[86],"outputs":[88],"corresponding":[89],"answers.":[90,108],"We":[91],"hypothesize":[92],"combination":[95,136],"these":[97,138],"is":[100,141],"mechanism":[102,143,159],"used":[103],"to":[104],"produce":[105],"correct":[106],"test":[110],"this,":[111],"categorize":[113],"each":[114],"neuron":[115],"into":[116],"several":[117,176],"types-such":[119],"activate":[123],"when":[124],"an":[125],"operand":[126],"falls":[127],"within":[128],"certain":[130],"range-and":[131],"find":[132],"unordered":[135],"types":[140],"accuracy":[150,167],"prompts.":[153],"Finally,":[154],"demonstrate":[156],"appears":[160],"main":[163],"source":[164],"early":[168],"training.":[170],"Overall,":[171],"our":[172],"experimental":[173],"results":[174],"across":[175],"LLMs":[177,180],"show":[178],"perform":[181],"using":[183],"neither":[184],"algorithms":[186],"nor":[187],"memorization;":[188],"rather,":[189],"rely":[191],"\"bag":[194],"heuristics\".":[196]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":2}],"updated_date":"2026-10-16T09:59:22.235173","created_date":"2024-11-14T00:00:00"}
