{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2KZB3R3BRVSEAWT46WGPNFPZUI","short_pith_number":"pith:2KZB3R3B","schema_version":"1.0","canonical_sha256":"d2b21dc7618d64405a7cf58cf695f9a21e2aba2db568538f5453d7e7b3e048fc","source":{"kind":"arxiv","id":"2506.11702","version":1},"attestation_state":"computed","paper":{"title":"Configurable Preference Tuning with Rubric-Guided Synthetic Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"V\\'ictor Gallego","submitted_at":"2025-06-13T12:17:38Z","abstract_excerpt":"Models of human feedback for AI alignment, such as those underpinning Direct Preference Optimization (DPO), often bake in a singular, static set of preferences, limiting adaptability. This paper challenges the assumption of monolithic preferences by introducing Configurable Preference Tuning (CPT), a novel framework for endowing language models with the ability to dynamically adjust their behavior based on explicit, human-interpretable directives. CPT leverages synthetically generated preference data, conditioned on system prompts derived from structured, fine-grained rubrics that define desir"},"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":"2506.11702","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-13T12:17:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3cc9fe1b289c1cc3855a14f9cf0ee5dcd476fd76578ca1a7cdc72bf521aecd05","abstract_canon_sha256":"f298f8b5c524fb990a74271996abd9ae88e8293371bb463bb23faa40cdaf1913"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:21:09.421530Z","signature_b64":"gTXCM8JYsODZRGyRVA+RnYGR9ikjwNSl8Y7ZkSuqr1mWD6oxsA+rsW1QTEoWnVyVQIWPzcPVS9vRhgtzsHN3Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d2b21dc7618d64405a7cf58cf695f9a21e2aba2db568538f5453d7e7b3e048fc","last_reissued_at":"2026-07-05T11:21:09.421048Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:21:09.421048Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Configurable Preference Tuning with Rubric-Guided Synthetic Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"V\\'ictor Gallego","submitted_at":"2025-06-13T12:17:38Z","abstract_excerpt":"Models of human feedback for AI alignment, such as those underpinning Direct Preference Optimization (DPO), often bake in a singular, static set of preferences, limiting adaptability. This paper challenges the assumption of monolithic preferences by introducing Configurable Preference Tuning (CPT), a novel framework for endowing language models with the ability to dynamically adjust their behavior based on explicit, human-interpretable directives. CPT leverages synthetically generated preference data, conditioned on system prompts derived from structured, fine-grained rubrics that define desir"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.11702","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/2506.11702/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":"2506.11702","created_at":"2026-07-05T11:21:09.421109+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.11702v1","created_at":"2026-07-05T11:21:09.421109+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.11702","created_at":"2026-07-05T11:21:09.421109+00:00"},{"alias_kind":"pith_short_12","alias_value":"2KZB3R3BRVSE","created_at":"2026-07-05T11:21:09.421109+00:00"},{"alias_kind":"pith_short_16","alias_value":"2KZB3R3BRVSEAWT4","created_at":"2026-07-05T11:21:09.421109+00:00"},{"alias_kind":"pith_short_8","alias_value":"2KZB3R3B","created_at":"2026-07-05T11:21:09.421109+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.17746","citing_title":"Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI","json":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI.json","graph_json":"https://pith.science/api/pith-number/2KZB3R3BRVSEAWT46WGPNFPZUI/graph.json","events_json":"https://pith.science/api/pith-number/2KZB3R3BRVSEAWT46WGPNFPZUI/events.json","paper":"https://pith.science/paper/2KZB3R3B"},"agent_actions":{"view_html":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI","download_json":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI.json","view_paper":"https://pith.science/paper/2KZB3R3B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.11702&json=true","fetch_graph":"https://pith.science/api/pith-number/2KZB3R3BRVSEAWT46WGPNFPZUI/graph.json","fetch_events":"https://pith.science/api/pith-number/2KZB3R3BRVSEAWT46WGPNFPZUI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI/action/storage_attestation","attest_author":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI/action/author_attestation","sign_citation":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI/action/citation_signature","submit_replication":"https://pith.science/pith/2KZB3R3BRVSEAWT46WGPNFPZUI/action/replication_record"}},"created_at":"2026-07-05T11:21:09.421109+00:00","updated_at":"2026-07-05T11:21:09.421109+00:00"}