{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:B57AAHQISPJ4FRMPEPJVB76UPE","short_pith_number":"pith:B57AAHQI","schema_version":"1.0","canonical_sha256":"0f7e001e0893d3c2c58f23d350ffd479197ccb5c24805c0ac6f8e3d3f4194f59","source":{"kind":"arxiv","id":"2407.04181","version":1},"attestation_state":"computed","paper":{"title":"Orchestrating LLMs with Different Personalizations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Jason Yuan, Jingwen Gu, Jin Peng Zhou, Katie Z Luo, Kilian Q. Weinberger, Wen Sun","submitted_at":"2024-07-04T22:55:02Z","abstract_excerpt":"This paper presents a novel approach to aligning large language models (LLMs) with individual human preferences, sometimes referred to as Reinforcement Learning from \\textit{Personalized} Human Feedback (RLPHF). Given stated preferences along multiple dimensions, such as helpfulness, conciseness, or humor, the goal is to create an LLM without re-training that best adheres to this specification. Starting from specialized expert LLMs, each trained for one such particular preference dimension, we propose a black-box method that merges their outputs on a per-token level. We train a lightweight Pre"},"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":"2407.04181","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-07-04T22:55:02Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"7751ce7530ec001add7f24a318630472faa7e2a3ab73f43664b49a4fc2d7ccc3","abstract_canon_sha256":"05b99844c63467869fbddf8c2741096fe74e12388a0d923eb9a41b299eb6d11e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:40:30.138808Z","signature_b64":"pxH9nbo51qrz4djYou8JQKxXaR1eDXXMl9eHeIYnqq/FgZXxh0RVdtucIIMCx3UqcvXnENioktYWTAyxS14hBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0f7e001e0893d3c2c58f23d350ffd479197ccb5c24805c0ac6f8e3d3f4194f59","last_reissued_at":"2026-07-05T08:40:30.138347Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:40:30.138347Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Orchestrating LLMs with Different Personalizations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Jason Yuan, Jingwen Gu, Jin Peng Zhou, Katie Z Luo, Kilian Q. Weinberger, Wen Sun","submitted_at":"2024-07-04T22:55:02Z","abstract_excerpt":"This paper presents a novel approach to aligning large language models (LLMs) with individual human preferences, sometimes referred to as Reinforcement Learning from \\textit{Personalized} Human Feedback (RLPHF). Given stated preferences along multiple dimensions, such as helpfulness, conciseness, or humor, the goal is to create an LLM without re-training that best adheres to this specification. Starting from specialized expert LLMs, each trained for one such particular preference dimension, we propose a black-box method that merges their outputs on a per-token level. We train a lightweight Pre"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.04181","kind":"arxiv","version":1},"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/2407.04181/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":"2407.04181","created_at":"2026-07-05T08:40:30.138400+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.04181v1","created_at":"2026-07-05T08:40:30.138400+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.04181","created_at":"2026-07-05T08:40:30.138400+00:00"},{"alias_kind":"pith_short_12","alias_value":"B57AAHQISPJ4","created_at":"2026-07-05T08:40:30.138400+00:00"},{"alias_kind":"pith_short_16","alias_value":"B57AAHQISPJ4FRMP","created_at":"2026-07-05T08:40:30.138400+00:00"},{"alias_kind":"pith_short_8","alias_value":"B57AAHQI","created_at":"2026-07-05T08:40:30.138400+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.07162","citing_title":"CLIPer: Tailoring Diverse User Preference via Classifier-Guided Inference-Time Personalization","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE","json":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE.json","graph_json":"https://pith.science/api/pith-number/B57AAHQISPJ4FRMPEPJVB76UPE/graph.json","events_json":"https://pith.science/api/pith-number/B57AAHQISPJ4FRMPEPJVB76UPE/events.json","paper":"https://pith.science/paper/B57AAHQI"},"agent_actions":{"view_html":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE","download_json":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE.json","view_paper":"https://pith.science/paper/B57AAHQI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.04181&json=true","fetch_graph":"https://pith.science/api/pith-number/B57AAHQISPJ4FRMPEPJVB76UPE/graph.json","fetch_events":"https://pith.science/api/pith-number/B57AAHQISPJ4FRMPEPJVB76UPE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE/action/storage_attestation","attest_author":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE/action/author_attestation","sign_citation":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE/action/citation_signature","submit_replication":"https://pith.science/pith/B57AAHQISPJ4FRMPEPJVB76UPE/action/replication_record"}},"created_at":"2026-07-05T08:40:30.138400+00:00","updated_at":"2026-07-05T08:40:30.138400+00:00"}