{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,27]],"date-time":"2026-03-27T17:10:53Z","timestamp":1774631453615,"version":"3.50.1"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"1","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AIES"],"abstract":"<jats:p>Large language models (LLMs) have been shown to be\neffective on tabular prediction tasks in the low-data\nregime, leveraging their internal knowledge and ability to\nlearn from instructions and examples. However, LLMs can\nfail to generate predictions that satisfy group fairness,\nthat is, produce equitable outcomes across groups.\nCritically, conventional debiasing approaches for natural\nlanguage tasks do not directly translate to mitigating\ngroup unfairness in tabular settings. In this work, we\nsystematically investigate four empirical approaches to\nimprove group fairness of LLM predictions on tabular\ndatasets, including fair prompt optimization, soft prompt\ntuning, strategic selection of few-shot examples, and\nself-refining predictions via chain-of-thought reasoning.\nThrough experiments on four tabular datasets using both\nopen-source and proprietary LLMs, we show the effectiveness\nof these methods in enhancing demographic parity while\nmaintaining high overall performance. Our analysis provides\nactionable insights for practitioners in selecting the most\nsuitable approach based on their specific requirements and\nconstraints.<\/jats:p>","DOI":"10.1609\/aies.v8i1.36572","type":"journal-article","created":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:17:53Z","timestamp":1760534273000},"page":"579-590","source":"Crossref","is-referenced-by-count":1,"title":["Improving LLM Group Fairness on Tabular Data via In-Context Learning"],"prefix":"10.1609","volume":"8","author":[{"given":"Valeriia","family":"Cherepanova","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Chia-Jung","family":"Lee","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Nil-Jana","family":"Akpinar","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Riccardo","family":"Fogliato","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Martin","family":"Bertran Lopez","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Michael","family":"Kearns","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"James","family":"Zou","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"9382","published-online":{"date-parts":[[2025,10,15]]},"container-title":["Proceedings of the AAAI\/ACM Conference on AI, Ethics, and Society"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36572\/38710","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36572\/38710","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:17:53Z","timestamp":1760534273000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/view\/36572"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,15]]},"references-count":0,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,10,15]]}},"URL":"https:\/\/doi.org\/10.1609\/aies.v8i1.36572","relation":{},"ISSN":["3065-8365"],"issn-type":[{"value":"3065-8365","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,15]]}}}