{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CMK5435CCMC45FU53QYSU2TW4U","short_pith_number":"pith:CMK5435C","schema_version":"1.0","canonical_sha256":"1315de6fa21305ce969ddc312a6a76e52f2c16d2d9ccc135189808cf4931448d","source":{"kind":"arxiv","id":"2412.15650","version":1},"attestation_state":"computed","paper":{"title":"Beyond Human Data: Aligning Multimodal Large Language Models by Iterative Self-Evolution","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Changxing Ding, Chao Xue, Qiong Cao, Wentao Tan, Yibing Zhan","submitted_at":"2024-12-20T08:06:00Z","abstract_excerpt":"Human preference alignment can greatly enhance Multimodal Large Language Models (MLLMs), but collecting high-quality preference data is costly. A promising solution is the self-evolution strategy, where models are iteratively trained on data they generate. However, current techniques still rely on human- or GPT-annotated data and sometimes require additional models or ground truth answers. To address these issues, we propose a novel multimodal self-evolution framework that enables the model to autonomously generate high-quality questions and answers using only unannotated images.\n  First, we i"},"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":"2412.15650","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-20T08:06:00Z","cross_cats_sorted":[],"title_canon_sha256":"4ab00beb2391238118abc0349ea63485fddd9c88da7b455bc9773a520f22d901","abstract_canon_sha256":"73d6e35563f95d2487b059b5ebf98a7afe8b87f8444fdd31ccd4faacc0fe1d9e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:52:27.716593Z","signature_b64":"LGP2kzlx1x/1lZcTg0y5N/itqt4xA7U9lJFFWhOZQFkOKfgVh3ZqJ0Nfj3GUBGt7id702pLXxGtcei/BmO8PDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1315de6fa21305ce969ddc312a6a76e52f2c16d2d9ccc135189808cf4931448d","last_reissued_at":"2026-07-05T09:52:27.716102Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:52:27.716102Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Human Data: Aligning Multimodal Large Language Models by Iterative Self-Evolution","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Changxing Ding, Chao Xue, Qiong Cao, Wentao Tan, Yibing Zhan","submitted_at":"2024-12-20T08:06:00Z","abstract_excerpt":"Human preference alignment can greatly enhance Multimodal Large Language Models (MLLMs), but collecting high-quality preference data is costly. A promising solution is the self-evolution strategy, where models are iteratively trained on data they generate. However, current techniques still rely on human- or GPT-annotated data and sometimes require additional models or ground truth answers. To address these issues, we propose a novel multimodal self-evolution framework that enables the model to autonomously generate high-quality questions and answers using only unannotated images.\n  First, we i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.15650","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/2412.15650/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":"2412.15650","created_at":"2026-07-05T09:52:27.716162+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.15650v1","created_at":"2026-07-05T09:52:27.716162+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.15650","created_at":"2026-07-05T09:52:27.716162+00:00"},{"alias_kind":"pith_short_12","alias_value":"CMK5435CCMC4","created_at":"2026-07-05T09:52:27.716162+00:00"},{"alias_kind":"pith_short_16","alias_value":"CMK5435CCMC45FU5","created_at":"2026-07-05T09:52:27.716162+00:00"},{"alias_kind":"pith_short_8","alias_value":"CMK5435C","created_at":"2026-07-05T09:52:27.716162+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.01048","citing_title":"Compared to What? Baselines and Metrics for Counterfactual Prompting","ref_index":72,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U","json":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U.json","graph_json":"https://pith.science/api/pith-number/CMK5435CCMC45FU53QYSU2TW4U/graph.json","events_json":"https://pith.science/api/pith-number/CMK5435CCMC45FU53QYSU2TW4U/events.json","paper":"https://pith.science/paper/CMK5435C"},"agent_actions":{"view_html":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U","download_json":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U.json","view_paper":"https://pith.science/paper/CMK5435C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.15650&json=true","fetch_graph":"https://pith.science/api/pith-number/CMK5435CCMC45FU53QYSU2TW4U/graph.json","fetch_events":"https://pith.science/api/pith-number/CMK5435CCMC45FU53QYSU2TW4U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U/action/storage_attestation","attest_author":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U/action/author_attestation","sign_citation":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U/action/citation_signature","submit_replication":"https://pith.science/pith/CMK5435CCMC45FU53QYSU2TW4U/action/replication_record"}},"created_at":"2026-07-05T09:52:27.716162+00:00","updated_at":"2026-07-05T09:52:27.716162+00:00"}