{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SO4HV3VJLYVPMLNGZFXQF6IHG3","short_pith_number":"pith:SO4HV3VJ","schema_version":"1.0","canonical_sha256":"93b87aeea95e2af62da6c96f02f90736f6024fb7cea43a8acde22b5fa8e92941","source":{"kind":"arxiv","id":"2503.16430","version":3},"attestation_state":"computed","paper":{"title":"Bridging Continuous and Discrete Tokens for Autoregressive Visual Generation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jiashi Feng, Shuhuai Ren, Xihui Liu, Yao Teng, Yuanzhi Zhu, Yuqing Wang, Zhijie Lin","submitted_at":"2025-03-20T17:59:59Z","abstract_excerpt":"Autoregressive visual generation models typically rely on tokenizers to compress images into tokens that can be predicted sequentially. A fundamental dilemma exists in token representation: discrete tokens enable straightforward modeling with standard cross-entropy loss, but suffer from information loss and tokenizer training instability; continuous tokens better preserve visual details, but require complex distribution modeling, complicating the generation pipeline. In this paper, we propose TokenBridge, which bridges this gap by maintaining the strong representation capacity of continuous to"},"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":"2503.16430","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-20T17:59:59Z","cross_cats_sorted":[],"title_canon_sha256":"d260e7dd2bbe8e3ed26d8f48f36317255dee0fcc27b60fbf5b2fd308708e3229","abstract_canon_sha256":"9f3844c9b3e3b5fe1f54ad406c22a4a4bddfe0150a38d2c2106ac440e92595e6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:01:21.445805Z","signature_b64":"TBnnggDjKRAhTUJIeY2ooytFS+bgfF0din+Vlh2FxUaqIvBn/j2RpQZyctWAzr3xeNeJaA2ZI/0QGuaEAdQGDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93b87aeea95e2af62da6c96f02f90736f6024fb7cea43a8acde22b5fa8e92941","last_reissued_at":"2026-07-05T12:01:21.445195Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:01:21.445195Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Bridging Continuous and Discrete Tokens for Autoregressive Visual Generation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jiashi Feng, Shuhuai Ren, Xihui Liu, Yao Teng, Yuanzhi Zhu, Yuqing Wang, Zhijie Lin","submitted_at":"2025-03-20T17:59:59Z","abstract_excerpt":"Autoregressive visual generation models typically rely on tokenizers to compress images into tokens that can be predicted sequentially. A fundamental dilemma exists in token representation: discrete tokens enable straightforward modeling with standard cross-entropy loss, but suffer from information loss and tokenizer training instability; continuous tokens better preserve visual details, but require complex distribution modeling, complicating the generation pipeline. In this paper, we propose TokenBridge, which bridges this gap by maintaining the strong representation capacity of continuous to"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.16430","kind":"arxiv","version":3},"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/2503.16430/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":"2503.16430","created_at":"2026-07-05T12:01:21.445251+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.16430v3","created_at":"2026-07-05T12:01:21.445251+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.16430","created_at":"2026-07-05T12:01:21.445251+00:00"},{"alias_kind":"pith_short_12","alias_value":"SO4HV3VJLYVP","created_at":"2026-07-05T12:01:21.445251+00:00"},{"alias_kind":"pith_short_16","alias_value":"SO4HV3VJLYVPMLNG","created_at":"2026-07-05T12:01:21.445251+00:00"},{"alias_kind":"pith_short_8","alias_value":"SO4HV3VJ","created_at":"2026-07-05T12:01:21.445251+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11363","citing_title":"NSVQ: Mitigating Codebook Collapse by Stabilizing Encoder Drift in Vector Quantization","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06870","citing_title":"Continuous First, Discrete Later: VQ-VAEs Without Dimensional Collapse","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08029","citing_title":"STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06870","citing_title":"Continuous First, Discrete Later: VQ-VAEs Without Dimensional Collapse","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3","json":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3.json","graph_json":"https://pith.science/api/pith-number/SO4HV3VJLYVPMLNGZFXQF6IHG3/graph.json","events_json":"https://pith.science/api/pith-number/SO4HV3VJLYVPMLNGZFXQF6IHG3/events.json","paper":"https://pith.science/paper/SO4HV3VJ"},"agent_actions":{"view_html":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3","download_json":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3.json","view_paper":"https://pith.science/paper/SO4HV3VJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.16430&json=true","fetch_graph":"https://pith.science/api/pith-number/SO4HV3VJLYVPMLNGZFXQF6IHG3/graph.json","fetch_events":"https://pith.science/api/pith-number/SO4HV3VJLYVPMLNGZFXQF6IHG3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3/action/storage_attestation","attest_author":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3/action/author_attestation","sign_citation":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3/action/citation_signature","submit_replication":"https://pith.science/pith/SO4HV3VJLYVPMLNGZFXQF6IHG3/action/replication_record"}},"created_at":"2026-07-05T12:01:21.445251+00:00","updated_at":"2026-07-05T12:01:21.445251+00:00"}