{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:CHLSQ2OOEZL3LSZTTAXEIPZEHU","short_pith_number":"pith:CHLSQ2OO","schema_version":"1.0","canonical_sha256":"11d72869ce2657b5cb33982e443f243d3fc4f806b538b5cb90d3f78b691e3949","source":{"kind":"arxiv","id":"2605.18395","version":1},"attestation_state":"computed","paper":{"title":"Diagnosing Korean-Language LLM Political Bias via Census-Grounded Agent Simulation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Census-grounded agent simulations diagnose political biases in Korean LLMs and accurately predict real election outcomes.","cross_cats":["cs.AI"],"primary_cat":"cs.CY","authors_text":"Sungwoo Kang","submitted_at":"2026-05-18T13:42:23Z","abstract_excerpt":"Large language models (LLMs) exhibit systematic political biases in voter simulations, but their underlying mechanisms and cross-lingual generalizations remain poorly understood. We introduce Dynamo-K, a census-grounded simulation framework evaluating Korean-language LLM political behavior across four models on six Korean elections (2017-2025). Using this framework, we identify three systematic failure modes: (1) progressive bias in moderate agents, where explicit mitigation reduces Mean Absolute Error (MAE) by 5.2 times; (2) model-dependent third-party salience collapse, distinguishing betwee"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.18395","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CY","submitted_at":"2026-05-18T13:42:23Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"63ae9df0108577604fcc866616e6d8d8bde6fc79acdb9cb68a1ae9126537009f","abstract_canon_sha256":"98f74b1801f553bf8e9dd8e336c256def41a25c7e608dbb0daa1f48ecc03c03e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:58.617441Z","signature_b64":"tznVidMpcYhJR2owUjU+zfXncaDHdQ9Juq3OxTg5iZ20+7d6emBzqWYa/ZQ+/EmZleP9daaQ6hIsZVdq6W1RCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"11d72869ce2657b5cb33982e443f243d3fc4f806b538b5cb90d3f78b691e3949","last_reissued_at":"2026-05-20T00:05:58.616595Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:58.616595Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Diagnosing Korean-Language LLM Political Bias via Census-Grounded Agent Simulation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Census-grounded agent simulations diagnose political biases in Korean LLMs and accurately predict real election outcomes.","cross_cats":["cs.AI"],"primary_cat":"cs.CY","authors_text":"Sungwoo Kang","submitted_at":"2026-05-18T13:42:23Z","abstract_excerpt":"Large language models (LLMs) exhibit systematic political biases in voter simulations, but their underlying mechanisms and cross-lingual generalizations remain poorly understood. We introduce Dynamo-K, a census-grounded simulation framework evaluating Korean-language LLM political behavior across four models on six Korean elections (2017-2025). Using this framework, we identify three systematic failure modes: (1) progressive bias in moderate agents, where explicit mitigation reduces Mean Absolute Error (MAE) by 5.2 times; (2) model-dependent third-party salience collapse, distinguishing betwee"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Dynamo-K accurately predicts 3/3 presidential winners including a 2.1%p MAE on the 2022 race with 0.73%p margin and correctly identifies the dominant party in a held-out local election.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that census-grounded agent profiles and scenario prompts produce voter behavior distributions that are sufficiently representative of actual Korean electorate responses for the purpose of diagnosing model bias.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Dynamo-K is a census-grounded agent simulation that diagnoses three failure modes in Korean LLMs' political predictions and shows calibration methods that improve accuracy on historical elections.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Census-grounded agent simulations diagnose political biases in Korean LLMs and accurately predict real election outcomes.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"57d98320053fdc2886d69f35331bb58df11be85062a97fd82d8427d4e54f5842"},"source":{"id":"2605.18395","kind":"arxiv","version":1},"verdict":{"id":"5b578aa0-c5dd-40dd-a10c-1569634e819b","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T23:59:58.839206Z","strongest_claim":"Dynamo-K accurately predicts 3/3 presidential winners including a 2.1%p MAE on the 2022 race with 0.73%p margin and correctly identifies the dominant party in a held-out local election.","one_line_summary":"Dynamo-K is a census-grounded agent simulation that diagnoses three failure modes in Korean LLMs' political predictions and shows calibration methods that improve accuracy on historical elections.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that census-grounded agent profiles and scenario prompts produce voter behavior distributions that are sufficiently representative of actual Korean electorate responses for the purpose of diagnosing model bias.","pith_extraction_headline":"Census-grounded agent simulations diagnose political biases in Korean LLMs and accurately predict real election outcomes."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.18395/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"cited_work_retraction","ran_at":"2026-05-19T23:52:10.869245Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T23:50:03.292214Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T23:33:29.753183Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"external_links","ran_at":"2026-05-19T23:31:43.709662Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T23:21:58.731675Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"03335653e9ddc7aca798710bc4b3b103fa45494c82f185060ecb24cce75fba1c"},"references":{"count":19,"sample":[{"doi":"","year":2026,"title":"Dynamo: Large-scale political simulation with LLM agents","work_id":"","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"L. P. Argyle, E. C. Busby, N. Fulda, J. R. Gubler, C. Rytting, and D. Wingate. Out of one, many: Using language models to simulate human samples.Political Analysis, 31(3):337–351, 2023","work_id":"","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1017/jea.2024.11","year":2022,"title":"M. D. Jenkins and H. J. Kim. The role of misogyny in the 2022 Korean presidential election: Under- standing the backlash against feminism in industrialized democracies.Journal of East Asian Studies, 2","work_id":"","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"S. Feng, C. Y. Park, Y. Liu, and Y. Tsvetkov. From pretraining data to language models to downstream tasks: Tracking the trails of political biases leading to unfair NLP models. InProceedings of the 6","work_id":"","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2000,"title":"G. Deffuant, D. Neau, F. Amblard, and G. Weisbuch. Mixing beliefs among interacting agents.Advances in Complex Systems, 3(01n04):87–98, 2000","work_id":"","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":0,"snapshot_sha256":"131bea81c46cd6deb1d8e6eb544a42a2f7206c999263acf3e064fb333cb1d49a","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.18395","created_at":"2026-05-20T00:05:58.616706+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.18395v1","created_at":"2026-05-20T00:05:58.616706+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18395","created_at":"2026-05-20T00:05:58.616706+00:00"},{"alias_kind":"pith_short_12","alias_value":"CHLSQ2OOEZL3","created_at":"2026-05-20T00:05:58.616706+00:00"},{"alias_kind":"pith_short_16","alias_value":"CHLSQ2OOEZL3LSZT","created_at":"2026-05-20T00:05:58.616706+00:00"},{"alias_kind":"pith_short_8","alias_value":"CHLSQ2OO","created_at":"2026-05-20T00:05:58.616706+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU","json":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU.json","graph_json":"https://pith.science/api/pith-number/CHLSQ2OOEZL3LSZTTAXEIPZEHU/graph.json","events_json":"https://pith.science/api/pith-number/CHLSQ2OOEZL3LSZTTAXEIPZEHU/events.json","paper":"https://pith.science/paper/CHLSQ2OO"},"agent_actions":{"view_html":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU","download_json":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU.json","view_paper":"https://pith.science/paper/CHLSQ2OO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.18395&json=true","fetch_graph":"https://pith.science/api/pith-number/CHLSQ2OOEZL3LSZTTAXEIPZEHU/graph.json","fetch_events":"https://pith.science/api/pith-number/CHLSQ2OOEZL3LSZTTAXEIPZEHU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU/action/storage_attestation","attest_author":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU/action/author_attestation","sign_citation":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU/action/citation_signature","submit_replication":"https://pith.science/pith/CHLSQ2OOEZL3LSZTTAXEIPZEHU/action/replication_record"}},"created_at":"2026-05-20T00:05:58.616706+00:00","updated_at":"2026-05-20T00:05:58.616706+00:00"}