{"paper":{"title":"Training-Free Cultural Alignment of Large Language Models via Persona Disagreement","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Disagreement among World Values Survey personas steers black-box LLMs toward country-specific cultural preferences at inference time.","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.CL","authors_text":"Chi-Nguyen Tran, Dao Sy Duy Minh, Huynh Trung Kiet, Long Tran-Thanh, Nguyen Lam Phu Quy, Phu-Hoa Pham, The Anh Han, Tuan Nguyen","submitted_at":"2026-05-11T16:55:16Z","abstract_excerpt":"Large language models increasingly mediate decisions that turn on moral judgement, yet a growing body of evidence shows that their implicit preferences are not culturally neutral. Existing cultural alignment methods either require per-country preference data and fine-tuning budgets or assume white-box access to model internals that commercial APIs do not expose. In this work, we focus on this realistic black-box, public-data-only regime and observe that within-country sociodemographic disagreement, not consensus, is the primary steering signal. We introduce DISCA (Disagreement-Informed Steerin"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across 20 countries and 7 open-weight backbones (2B--70B), DISCA reduces cultural misalignment on MultiTP by 10--24% on the six backbones >=3.8B, and 2--7% on open-ended scenarios, without changing any weights.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That within-country sociodemographic disagreement, when instantiated via World-Values-Survey-grounded persona agents, constitutes the primary and sufficient steering signal for correcting cultural misalignment in black-box LLMs.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DISCA uses disagreement among WVS-grounded persona panels to apply loss-averse logit corrections that reduce cultural misalignment by 10-24% on MultiTP for models 3.8B and larger, without weight changes.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Disagreement among World Values Survey personas steers black-box LLMs toward country-specific cultural preferences at inference time.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c7f3dbc709f0fd466893a41c27b25030bf211d590c62368266c239937b4b5d1b"},"source":{"id":"2605.10843","kind":"arxiv","version":2},"verdict":{"id":"f6d3b266-fadc-423a-bf13-bcc561037504","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T04:03:34.969346Z","strongest_claim":"Across 20 countries and 7 open-weight backbones (2B--70B), DISCA reduces cultural misalignment on MultiTP by 10--24% on the six backbones >=3.8B, and 2--7% on open-ended scenarios, without changing any weights.","one_line_summary":"DISCA uses disagreement among WVS-grounded persona panels to apply loss-averse logit corrections that reduce cultural misalignment by 10-24% on MultiTP for models 3.8B and larger, without weight changes.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That within-country sociodemographic disagreement, when instantiated via World-Values-Survey-grounded persona agents, constitutes the primary and sufficient steering signal for correcting cultural misalignment in black-box LLMs.","pith_extraction_headline":"Disagreement among World Values Survey personas steers black-box LLMs toward country-specific cultural preferences at inference time."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.10843/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T14:34:25.361729Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T10:31:17.530567Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T08:57:02.033032Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"532cc7b77c859253781ddaa0f3bc4fbe244c0783d1cfa5e0129e8ed94a582958"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"8a3ad2b99542d180c0ec73fd9aa1f770bf97c7540399a1f50c0461eaef155bd6"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}