A formal approach for understanding the behavior of constrained language models
Resumen:
We study language models under explicit formal constraints by treating common operations, such as prompting, masking, temperature scaling, and vocabulary translation, as composable transformations, in particular those that preserve the set of generable sequences and/or its probability distribution. We focus on constraints that disable tokens, potentially leading to dead-ends during generation. Moreover, we formalize bounded-length and safe decoding procedures. The theoretical results enable the algorithmic approximation of the probability of subsets (properties) of the generable language, as well as active learning of automata representations, such as DFA o Moore machines. We evaluate the approach to analyze the behavior of several state-of-the-art large language models.
| 2026 | |
| Agencia Nacional de Investigación e Innovación | |
|
Artificial Intelligence Language Models Constrained Decoding Formal Languages Ciencias Naturales y Exactas Ciencias de la Computación e Información |
|
| Inglés | |
| Agencia Nacional de Investigación e Innovación | |
| REDI | |
| https://hdl.handle.net/20.500.12381/5635 | |
| Acceso abierto | |
| Reconocimiento 4.0 Internacional. (CC BY) |
| _version_ | 1874662862400520192 |
|---|---|
| author | Iturbide, Martín |
| author2 | Yovine, Sergio Carrasco, Matías |
| author2_role | author author |
| author_facet | Iturbide, Martín Yovine, Sergio Carrasco, Matías |
| author_role | author |
| bitstream.checksum.fl_str_mv | a4ce09f01b5dd771727aa05c73851623 27b755c8281e38b04e3da3389076b036 |
| bitstream.checksumAlgorithm.fl_str_mv | MD5 MD5 |
| bitstream.url.fl_str_mv | https://redi.anii.org.uy/jspui/bitstream/20.500.12381/5635/2/license.txt https://redi.anii.org.uy/jspui/bitstream/20.500.12381/5635/1/ICGI_2026__bis___arXiv_.pdf |
| collection | REDI |
| dc.creator.none.fl_str_mv | Iturbide, Martín Yovine, Sergio Carrasco, Matías |
| dc.date.accessioned.none.fl_str_mv | 2026-08-17T18:35:39Z |
| dc.date.available.none.fl_str_mv | 2026-08-17T18:35:39Z |
| dc.date.issued.none.fl_str_mv | 2026-08 |
| dc.description.abstract.none.fl_txt_mv | We study language models under explicit formal constraints by treating common operations, such as prompting, masking, temperature scaling, and vocabulary translation, as composable transformations, in particular those that preserve the set of generable sequences and/or its probability distribution. We focus on constraints that disable tokens, potentially leading to dead-ends during generation. Moreover, we formalize bounded-length and safe decoding procedures. The theoretical results enable the algorithmic approximation of the probability of subsets (properties) of the generable language, as well as active learning of automata representations, such as DFA o Moore machines. We evaluate the approach to analyze the behavior of several state-of-the-art large language models. |
| dc.description.sponsorship.none.fl_txt_mv | Agencia Nacional de Investigación e Innovación |
| dc.identifier.anii.es.fl_str_mv | FMV_1_2023_1_175864 POS_FMV_2023_1_1012218 |
| dc.identifier.uri.none.fl_str_mv | https://hdl.handle.net/20.500.12381/5635 |
| dc.language.iso.none.fl_str_mv | eng |
| dc.relation.none.fl_str_mv | https://hdl.handle.net/20.500.12381/3417 https://hdl.handle.net/20.500.12381/3418 https://hdl.handle.net/20.500.12381/3419 https://hdl.handle.net/20.500.12381/3420 https://hdl.handle.net/20.500.12381/3622 https://hdl.handle.net/20.500.12381/3624 https://hdl.handle.net/20.500.12381/3626 https://hdl.handle.net/20.500.12381/3656 https://hdl.handle.net/20.500.12381/5138 https://doi.org/10.60895/redata/Z8QDEZ https://doi.org/10.60895/redata/NDHQQQ https://doi.org/10.60895/redata/KNERSJ https://doi.org/10.60895/redata/JY5DUS https://hdl.handle.net/20.500.12381/5632 https://hdl.handle.net/20.500.12381/5633 |
| dc.rights.*.fl_str_mv | Acceso abierto |
| dc.rights.license.none.fl_str_mv | Reconocimiento 4.0 Internacional. (CC BY) |
| dc.rights.none.fl_str_mv | info:eu-repo/semantics/openAccess |
| dc.source.none.fl_str_mv | reponame:REDI instname:Agencia Nacional de Investigación e Innovación instacron:Agencia Nacional de Investigación e Innovación |
| dc.subject.anii.none.fl_str_mv | Ciencias Naturales y Exactas Ciencias de la Computación e Información |
| dc.subject.es.fl_str_mv | Artificial Intelligence Language Models Constrained Decoding Formal Languages |
| dc.title.none.fl_str_mv | A formal approach for understanding the behavior of constrained language models |
| dc.type.es.fl_str_mv | Preprint |
| dc.type.none.fl_str_mv | info:eu-repo/semantics/preprint |
| description | We study language models under explicit formal constraints by treating common operations, such as prompting, masking, temperature scaling, and vocabulary translation, as composable transformations, in particular those that preserve the set of generable sequences and/or its probability distribution. We focus on constraints that disable tokens, potentially leading to dead-ends during generation. Moreover, we formalize bounded-length and safe decoding procedures. The theoretical results enable the algorithmic approximation of the probability of subsets (properties) of the generable language, as well as active learning of automata representations, such as DFA o Moore machines. We evaluate the approach to analyze the behavior of several state-of-the-art large language models. |
| eu_rights_str_mv | openAccess |
| format | preprint |
| id | REDI_bff03944a12f7528b278bff7ac45f7c4 |
| identifier_str_mv | FMV_1_2023_1_175864 POS_FMV_2023_1_1012218 |
| instacron_str | Agencia Nacional de Investigación e Innovación |
| institution | Agencia Nacional de Investigación e Innovación |
| instname_str | Agencia Nacional de Investigación e Innovación |
| language | eng |
| network_acronym_str | REDI |
| network_name_str | REDI |
| oai_identifier_str | oai:redi.anii.org.uy:20.500.12381/5635 |
| publishDate | 2026 |
| reponame_str | REDI |
| repository.mail.fl_str_mv | jmaldini@anii.org.uy |
| repository.name.fl_str_mv | REDI - Agencia Nacional de Investigación e Innovación |
| repository_id_str | 9421 |
| rights_invalid_str_mv | Reconocimiento 4.0 Internacional. (CC BY) Acceso abierto |
| spelling | Reconocimiento 4.0 Internacional. (CC BY)Acceso abiertoinfo:eu-repo/semantics/openAccess2026-08-17T18:35:39Z2026-08-17T18:35:39Z2026-08https://hdl.handle.net/20.500.12381/5635FMV_1_2023_1_175864POS_FMV_2023_1_1012218We study language models under explicit formal constraints by treating common operations, such as prompting, masking, temperature scaling, and vocabulary translation, as composable transformations, in particular those that preserve the set of generable sequences and/or its probability distribution. We focus on constraints that disable tokens, potentially leading to dead-ends during generation. Moreover, we formalize bounded-length and safe decoding procedures. The theoretical results enable the algorithmic approximation of the probability of subsets (properties) of the generable language, as well as active learning of automata representations, such as DFA o Moore machines. We evaluate the approach to analyze the behavior of several state-of-the-art large language models.Agencia Nacional de Investigación e Innovaciónenghttps://hdl.handle.net/20.500.12381/3417https://hdl.handle.net/20.500.12381/3418https://hdl.handle.net/20.500.12381/3419https://hdl.handle.net/20.500.12381/3420https://hdl.handle.net/20.500.12381/3622https://hdl.handle.net/20.500.12381/3624https://hdl.handle.net/20.500.12381/3626https://hdl.handle.net/20.500.12381/3656https://hdl.handle.net/20.500.12381/5138https://doi.org/10.60895/redata/Z8QDEZhttps://doi.org/10.60895/redata/NDHQQQhttps://doi.org/10.60895/redata/KNERSJhttps://doi.org/10.60895/redata/JY5DUShttps://hdl.handle.net/20.500.12381/5632https://hdl.handle.net/20.500.12381/5633Artificial IntelligenceLanguage ModelsConstrained DecodingFormal LanguagesCiencias Naturales y ExactasCiencias de la Computación e InformaciónA formal approach for understanding the behavior of constrained language modelsPreprintinfo:eu-repo/semantics/preprintUniversidad ORT Uruguay//Ciencias Naturales y Exactas/Ciencias de la Computación e Informaciónreponame:REDIinstname:Agencia Nacional de Investigación e Innovacióninstacron:Agencia Nacional de Investigación e InnovaciónIturbide, MartínYovine, SergioCarrasco, MatíasLICENSElicense.txtlicense.txttext/plain; charset=utf-84967https://redi.anii.org.uy/jspui/bitstream/20.500.12381/5635/2/license.txta4ce09f01b5dd771727aa05c73851623MD52ORIGINALICGI_2026__bis___arXiv_.pdfICGI_2026__bis___arXiv_.pdfapplication/pdf497181https://redi.anii.org.uy/jspui/bitstream/20.500.12381/5635/1/ICGI_2026__bis___arXiv_.pdf27b755c8281e38b04e3da3389076b036MD5120.500.12381/56352026-08-17 15:35:40.311oai:redi.anii.org.uy:20.500.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Institucionalhttps://redi.anii.org.uy/Gobiernohttps://www.anii.org.uy/https://redi.anii.org.uy/oai/requestjmaldini@anii.org.uyUruguayopendoar:94212026-08-17T18:35:40REDI - Agencia Nacional de Investigación e Innovaciónfalse |
| spellingShingle | A formal approach for understanding the behavior of constrained language models Iturbide, Martín Artificial Intelligence Language Models Constrained Decoding Formal Languages Ciencias Naturales y Exactas Ciencias de la Computación e Información |
| title | A formal approach for understanding the behavior of constrained language models |
| title_full | A formal approach for understanding the behavior of constrained language models |
| title_fullStr | A formal approach for understanding the behavior of constrained language models |
| title_full_unstemmed | A formal approach for understanding the behavior of constrained language models |
| title_short | A formal approach for understanding the behavior of constrained language models |
| title_sort | A formal approach for understanding the behavior of constrained language models |
| topic | Artificial Intelligence Language Models Constrained Decoding Formal Languages Ciencias Naturales y Exactas Ciencias de la Computación e Información |
| url | https://hdl.handle.net/20.500.12381/5635 |
