{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7YQEJHGH2JDVI3ANRDLVJEOXPG","short_pith_number":"pith:7YQEJHGH","schema_version":"1.0","canonical_sha256":"fe20449cc7d247546c0d88d75491d779a0c8c161aaaa50e3f7e1a13ebff1a6c1","source":{"kind":"arxiv","id":"2604.22280","version":2},"attestation_state":"computed","paper":{"title":"Beyond Chain-of-Thought: Rewrite as a Universal Interface for Generative Multimodal Embeddings","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Rewrite replaces chain-of-thought to create stronger generative multimodal embeddings.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bo Lin, Bosong Chai, Chenxi Zhao, Dacheng Yin, Feipeng Ma, Fengyun Rao, Hebei Li, Jie Chen, Jing Lyu, Junjie Zhou, Ke Mei, Peixi Wu, Shannan Yan, Tianyi Wang, Xiaoyan Sun, Yansong Peng, Zhangchi Hu, Zhibin Lan","submitted_at":"2026-04-24T06:50:11Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have emerged as a promising foundation for universal multimodal embeddings. Recent studies have shown that reasoning-driven generative multimodal embeddings can outperform discriminative embeddings on several embedding tasks. However, Chain-of-Thought (CoT) reasoning tends to generate redundant thinking steps and introduce semantic ambiguity in the summarized answers in broader retrieval scenarios. To address this limitation, we propose Rewrite-driven Multimodal Embedding (RIME), a unified framework that jointly optimizes generation and embedding throug"},"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":"2604.22280","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-24T06:50:11Z","cross_cats_sorted":[],"title_canon_sha256":"b647ff027d2017048bdb103922a95a827802218f096be6b1a85619d12a05d18b","abstract_canon_sha256":"aaf8d01b8f3650ec3080d517bdcdb44a80cc8d69f707cdf9ec82041367bec118"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T02:05:05.580700Z","signature_b64":"pTIN6PtXcL4Vfh+w4xn6uAllS0DlZHOb6pzAkOxFIdJymC90ODNsdUoSzfM5hjp35/CJQVsvNpq9zTUu8n31DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fe20449cc7d247546c0d88d75491d779a0c8c161aaaa50e3f7e1a13ebff1a6c1","last_reissued_at":"2026-05-29T02:05:05.579811Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T02:05:05.579811Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Chain-of-Thought: Rewrite as a Universal Interface for Generative Multimodal Embeddings","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Rewrite replaces chain-of-thought to create stronger generative multimodal embeddings.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bo Lin, Bosong Chai, Chenxi Zhao, Dacheng Yin, Feipeng Ma, Fengyun Rao, Hebei Li, Jie Chen, Jing Lyu, Junjie Zhou, Ke Mei, Peixi Wu, Shannan Yan, Tianyi Wang, Xiaoyan Sun, Yansong Peng, Zhangchi Hu, Zhibin Lan","submitted_at":"2026-04-24T06:50:11Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) have emerged as a promising foundation for universal multimodal embeddings. Recent studies have shown that reasoning-driven generative multimodal embeddings can outperform discriminative embeddings on several embedding tasks. However, Chain-of-Thought (CoT) reasoning tends to generate redundant thinking steps and introduce semantic ambiguity in the summarized answers in broader retrieval scenarios. To address this limitation, we propose Rewrite-driven Multimodal Embedding (RIME), a unified framework that jointly optimizes generation and embedding throug"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"RIME substantially outperforms prior generative embedding models while significantly reducing the length of thinking.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The rewrite step preserves necessary semantic information for downstream retrieval while remaining retrieval-friendly; this is asserted but not derived in the abstract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Rewrite-driven generation with alignment and RL produces shorter, more effective generative multimodal embeddings than CoT methods on retrieval benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Rewrite replaces chain-of-thought to create stronger generative multimodal embeddings.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"555391a6a64ccbb3cc505f71c65a52eb97be67a357f2dd78f5403972a224b678"},"source":{"id":"2604.22280","kind":"arxiv","version":2},"verdict":{"id":"086aaa4f-77c8-4688-a6a9-ba19cd8556d6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T12:43:13.536423Z","strongest_claim":"RIME substantially outperforms prior generative embedding models while significantly reducing the length of thinking.","one_line_summary":"Rewrite-driven generation with alignment and RL produces shorter, more effective generative multimodal embeddings than CoT methods on retrieval benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The rewrite step preserves necessary semantic information for downstream retrieval while remaining retrieval-friendly; this is asserted but not derived in the abstract.","pith_extraction_headline":"Rewrite replaces chain-of-thought to create stronger generative multimodal embeddings."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.22280/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T11:33:48.407352Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T00:06:50.576536Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"3c05b33fe034c3fc1777d4b403d3be20ab17c147cf1b21e100477e7196fb9334"},"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":"2604.22280","created_at":"2026-05-29T02:05:05.579923+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.22280v2","created_at":"2026-05-29T02:05:05.579923+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.22280","created_at":"2026-05-29T02:05:05.579923+00:00"},{"alias_kind":"pith_short_12","alias_value":"7YQEJHGH2JDV","created_at":"2026-05-29T02:05:05.579923+00:00"},{"alias_kind":"pith_short_16","alias_value":"7YQEJHGH2JDVI3AN","created_at":"2026-05-29T02:05:05.579923+00:00"},{"alias_kind":"pith_short_8","alias_value":"7YQEJHGH","created_at":"2026-05-29T02:05:05.579923+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/7YQEJHGH2JDVI3ANRDLVJEOXPG","json":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG.json","graph_json":"https://pith.science/api/pith-number/7YQEJHGH2JDVI3ANRDLVJEOXPG/graph.json","events_json":"https://pith.science/api/pith-number/7YQEJHGH2JDVI3ANRDLVJEOXPG/events.json","paper":"https://pith.science/paper/7YQEJHGH"},"agent_actions":{"view_html":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG","download_json":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG.json","view_paper":"https://pith.science/paper/7YQEJHGH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.22280&json=true","fetch_graph":"https://pith.science/api/pith-number/7YQEJHGH2JDVI3ANRDLVJEOXPG/graph.json","fetch_events":"https://pith.science/api/pith-number/7YQEJHGH2JDVI3ANRDLVJEOXPG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG/action/storage_attestation","attest_author":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG/action/author_attestation","sign_citation":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG/action/citation_signature","submit_replication":"https://pith.science/pith/7YQEJHGH2JDVI3ANRDLVJEOXPG/action/replication_record"}},"created_at":"2026-05-29T02:05:05.579923+00:00","updated_at":"2026-05-29T02:05:05.579923+00:00"}