{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:KF5QD37NZORVRRZVK5UQUOLGO3","short_pith_number":"pith:KF5QD37N","schema_version":"1.0","canonical_sha256":"517b01efedcba358c73557690a396676e56cd446f920d82851998812bfabee98","source":{"kind":"arxiv","id":"2204.08721","version":2},"attestation_state":"computed","paper":{"title":"Multimodal Token Fusion for Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fuchun Sun, Lele Cao, Wenbing Huang, Xinghao Chen, Yikai Wang, Yunhe Wang","submitted_at":"2022-04-19T07:47:50Z","abstract_excerpt":"Many adaptations of transformers have emerged to address the single-modal vision tasks, where self-attention modules are stacked to handle input sources like images. Intuitively, feeding multiple modalities of data to vision transformers could improve the performance, yet the inner-modal attentive weights may also be diluted, which could thus undermine the final performance. In this paper, we propose a multimodal token fusion method (TokenFusion), tailored for transformer-based vision tasks. To effectively fuse multiple modalities, TokenFusion dynamically detects uninformative tokens and subst"},"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":"2204.08721","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2022-04-19T07:47:50Z","cross_cats_sorted":[],"title_canon_sha256":"a7d4fbfbf6be53e4c64e363df0fdf1fa99b85cb118093cad8f74b37facc571bb","abstract_canon_sha256":"ca447869dfdf6650409d240b42fc7f15e978c2aa2dd4c8fc8d4deba63eab0f96"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:40:26.949130Z","signature_b64":"jgsbrHu5fRTMFpiPINxmt0sIRe4UROjH5j/zO6aFMbB6ktRHtrMYw+wnrREuvwP2q57oC52vdIsxViuqTLpVBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"517b01efedcba358c73557690a396676e56cd446f920d82851998812bfabee98","last_reissued_at":"2026-07-05T04:40:26.948662Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:40:26.948662Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multimodal Token Fusion for Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fuchun Sun, Lele Cao, Wenbing Huang, Xinghao Chen, Yikai Wang, Yunhe Wang","submitted_at":"2022-04-19T07:47:50Z","abstract_excerpt":"Many adaptations of transformers have emerged to address the single-modal vision tasks, where self-attention modules are stacked to handle input sources like images. Intuitively, feeding multiple modalities of data to vision transformers could improve the performance, yet the inner-modal attentive weights may also be diluted, which could thus undermine the final performance. In this paper, we propose a multimodal token fusion method (TokenFusion), tailored for transformer-based vision tasks. To effectively fuse multiple modalities, TokenFusion dynamically detects uninformative tokens and subst"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2204.08721","kind":"arxiv","version":2},"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/2204.08721/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":"2204.08721","created_at":"2026-07-05T04:40:26.948728+00:00"},{"alias_kind":"arxiv_version","alias_value":"2204.08721v2","created_at":"2026-07-05T04:40:26.948728+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2204.08721","created_at":"2026-07-05T04:40:26.948728+00:00"},{"alias_kind":"pith_short_12","alias_value":"KF5QD37NZORV","created_at":"2026-07-05T04:40:26.948728+00:00"},{"alias_kind":"pith_short_16","alias_value":"KF5QD37NZORVRRZV","created_at":"2026-07-05T04:40:26.948728+00:00"},{"alias_kind":"pith_short_8","alias_value":"KF5QD37N","created_at":"2026-07-05T04:40:26.948728+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.08096","citing_title":"THUD++: Large-Scale Dynamic Indoor Scene Dataset and Benchmark for Mobile Robots","ref_index":38,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3","json":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3.json","graph_json":"https://pith.science/api/pith-number/KF5QD37NZORVRRZVK5UQUOLGO3/graph.json","events_json":"https://pith.science/api/pith-number/KF5QD37NZORVRRZVK5UQUOLGO3/events.json","paper":"https://pith.science/paper/KF5QD37N"},"agent_actions":{"view_html":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3","download_json":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3.json","view_paper":"https://pith.science/paper/KF5QD37N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2204.08721&json=true","fetch_graph":"https://pith.science/api/pith-number/KF5QD37NZORVRRZVK5UQUOLGO3/graph.json","fetch_events":"https://pith.science/api/pith-number/KF5QD37NZORVRRZVK5UQUOLGO3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3/action/storage_attestation","attest_author":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3/action/author_attestation","sign_citation":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3/action/citation_signature","submit_replication":"https://pith.science/pith/KF5QD37NZORVRRZVK5UQUOLGO3/action/replication_record"}},"created_at":"2026-07-05T04:40:26.948728+00:00","updated_at":"2026-07-05T04:40:26.948728+00:00"}