{"id":"https://openalex.org/W4416033906","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.1109","title":"When Punctuation Matters: A Large-Scale Comparison of Prompt Robustness Methods for LLMs","display_name":"When Punctuation Matters: A Large-Scale Comparison of Prompt Robustness Methods for LLMs","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416033906","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.1109"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.1109","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.1109","pdf_url":"https://aclanthology.org/2025.findings-emnlp.1109.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-emnlp.1109.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5077094574","display_name":"Mikhail Seleznyov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mikhail Seleznyov","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5118713184","display_name":"Mikhail Chaichuk","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mikhail Chaichuk","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120308432","display_name":"Gleb Ershov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gleb Ershov","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026157285","display_name":"Alexander Panchenko","orcid":"https://orcid.org/0000-0001-6097-6118"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Alexander Panchenko","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012311258","display_name":"Elena Tutubalina","orcid":"https://orcid.org/0000-0001-7936-0284"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Elena Tutubalina","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5092596475","display_name":"Oleg Somov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oleg Somov","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":"20370","last_page":"20385"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12072","display_name":"Machine Learning and Algorithms","score":0.06800000369548798,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.06800000369548798,"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":0.061900001019239426,"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":0.05119999870657921,"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/punctuation","display_name":"Punctuation","score":0.7146999835968018},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.616100013256073}],"concepts":[{"id":"https://openalex.org/C540372491","wikidata":"https://www.wikidata.org/wiki/Q82622","display_name":"Punctuation","level":2,"score":0.7146999835968018},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.616100013256073},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5626999735832214},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48739999532699585},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3610999882221222},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2418999969959259},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.22439999878406525},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.20999999344348907},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.1891999989748001},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.18410000205039978}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.1109","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.1109","pdf_url":"https://aclanthology.org/2025.findings-emnlp.1109.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.1109","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.1109","pdf_url":"https://aclanthology.org/2025.findings-emnlp.1109.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320324261","display_name":"National Research University Higher School of Economics","ror":"https://ror.org/055f7t516"},{"id":"https://openalex.org/F4320324632","display_name":"Ministero dello Sviluppo Economico","ror":"https://ror.org/011z3ff80"},{"id":"https://openalex.org/F4320334801","display_name":"Russian Academy of Sciences","ror":"https://ror.org/05qrfxd25"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416033906.pdf","grobid_xml":"https://content.openalex.org/works/W4416033906.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"are":[4],"highly":[5],"sensitive":[6],"to":[7,80,85,91,108],"subtle,":[8],"non-semantic":[9],"variations":[10],"in":[11,120],"prompt":[12,29],"phrasing":[13],"and":[14,45,62,66,82,116],"formatting.In":[15],"this":[16],"work,":[17],"we":[18,76],"present":[19],"the":[20,99],"first":[21],"systematic":[22],"evaluation":[23,55],"of":[24,73,102],"5":[25],"methods":[26,58],"for":[27,114],"improving":[28],"robustness":[30,57,90,104],"within":[31],"a":[32],"unified":[33],"experimental":[34],"framework.We":[35],"benchmark":[36],"these":[37,103],"techniques":[38],"on":[39],"8":[40],"models":[41],"from":[42,51,59],"Llama,":[43],"Qwen":[44],"Gemma":[46],"families":[47],"across":[48],"52":[49],"tasks":[50],"Natural":[52],"Instructions":[53],"dataset.Our":[54],"covers":[56],"both":[60],"fine-tuned":[61],"in-context":[63],"learning":[64],"paradigms,":[65],"tests":[67],"their":[68],"generalization":[69],"against":[70],"multiple":[71],"types":[72],"distribution":[74],"shifts.Finally,":[75],"extend":[77],"our":[78],"analysis":[79],"GPT-4.1":[81],"DeepSeek":[83],"V3":[84],"assess":[86],"frontier":[87],"models'":[88],"current":[89],"format":[92],"perturbations.Our":[93],"findings":[94],"offer":[95],"actionable":[96],"insights":[97],"into":[98],"relative":[100],"effectiveness":[101],"methods,":[105],"enabling":[106],"practitioners":[107],"make":[109],"informed":[110],"decisions":[111],"when":[112],"aiming":[113],"stable":[115],"reliable":[117],"LLM":[118],"performance":[119],"real-world":[121],"applications.":[122]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-11-08T00:00:00"}