{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:7TZ3NQL5MQQ73VH3DRCUJZH52L","short_pith_number":"pith:7TZ3NQL5","schema_version":"1.0","canonical_sha256":"fcf3b6c17d6421fdd4fb1c4544e4fdd2d2c657ea8f4a1f9dddce5257d09e47b0","source":{"kind":"arxiv","id":"2402.15018","version":2},"attestation_state":"computed","paper":{"title":"Unintended Impacts of LLM Alignment on Global Representation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CY","cs.LG"],"primary_cat":"cs.CL","authors_text":"Diyi Yang, Michael J. Ryan, William Held","submitted_at":"2024-02-22T23:31:22Z","abstract_excerpt":"Before being deployed for user-facing applications, developers align Large Language Models (LLMs) to user preferences through a variety of procedures, such as Reinforcement Learning From Human Feedback (RLHF) and Direct Preference Optimization (DPO). Current evaluations of these procedures focus on benchmarks of instruction following, reasoning, and truthfulness. However, human preferences are not universal, and aligning to specific preference sets may have unintended effects. We explore how alignment impacts performance along three axes of global representation: English dialects, multilingual"},"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":"2402.15018","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-22T23:31:22Z","cross_cats_sorted":["cs.CY","cs.LG"],"title_canon_sha256":"4da9c71410e9e41c6600fb251de3a9c665e91ed3851268b022410d27fefe086b","abstract_canon_sha256":"d9b9716a9c47f4cfb193d90485381cba50fb6fcdad0f2c86e637dab867e05067"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:38.773024Z","signature_b64":"6AavqXYcUe2rS8ZXyA8705XiiA+82mt5Fr9c0APCu6lYnFPnjhjoX+fhDzcJmLr7LCuBkDG5HXbAF5UphJ9bDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fcf3b6c17d6421fdd4fb1c4544e4fdd2d2c657ea8f4a1f9dddce5257d09e47b0","last_reissued_at":"2026-07-05T08:28:38.772553Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:38.772553Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unintended Impacts of LLM Alignment on Global Representation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CY","cs.LG"],"primary_cat":"cs.CL","authors_text":"Diyi Yang, Michael J. Ryan, William Held","submitted_at":"2024-02-22T23:31:22Z","abstract_excerpt":"Before being deployed for user-facing applications, developers align Large Language Models (LLMs) to user preferences through a variety of procedures, such as Reinforcement Learning From Human Feedback (RLHF) and Direct Preference Optimization (DPO). Current evaluations of these procedures focus on benchmarks of instruction following, reasoning, and truthfulness. However, human preferences are not universal, and aligning to specific preference sets may have unintended effects. We explore how alignment impacts performance along three axes of global representation: English dialects, multilingual"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.15018","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2402.15018/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","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":"2402.15018","created_at":"2026-07-05T08:28:38.772615+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.15018v2","created_at":"2026-07-05T08:28:38.772615+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.15018","created_at":"2026-07-05T08:28:38.772615+00:00"},{"alias_kind":"pith_short_12","alias_value":"7TZ3NQL5MQQ7","created_at":"2026-07-05T08:28:38.772615+00:00"},{"alias_kind":"pith_short_16","alias_value":"7TZ3NQL5MQQ73VH3","created_at":"2026-07-05T08:28:38.772615+00:00"},{"alias_kind":"pith_short_8","alias_value":"7TZ3NQL5","created_at":"2026-07-05T08:28:38.772615+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2305.09620","citing_title":"AI-Augmented Surveys: Leveraging Large Language Models and Surveys for Opinion Prediction","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2412.20760","citing_title":"Attributing Culture-Conditioned Generations to Pretraining Corpora","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22749","citing_title":"Representational Harms in LLM-Generated Narratives Against Global Majority Nationalities","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02601","citing_title":"SemEval-2026 Task 7: Everyday Knowledge Across Diverse Languages and Cultures","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L","json":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L.json","graph_json":"https://pith.science/api/pith-number/7TZ3NQL5MQQ73VH3DRCUJZH52L/graph.json","events_json":"https://pith.science/api/pith-number/7TZ3NQL5MQQ73VH3DRCUJZH52L/events.json","paper":"https://pith.science/paper/7TZ3NQL5"},"agent_actions":{"view_html":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L","download_json":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L.json","view_paper":"https://pith.science/paper/7TZ3NQL5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.15018&json=true","fetch_graph":"https://pith.science/api/pith-number/7TZ3NQL5MQQ73VH3DRCUJZH52L/graph.json","fetch_events":"https://pith.science/api/pith-number/7TZ3NQL5MQQ73VH3DRCUJZH52L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L/action/storage_attestation","attest_author":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L/action/author_attestation","sign_citation":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L/action/citation_signature","submit_replication":"https://pith.science/pith/7TZ3NQL5MQQ73VH3DRCUJZH52L/action/replication_record"}},"created_at":"2026-07-05T08:28:38.772615+00:00","updated_at":"2026-07-05T08:28:38.772615+00:00"}