{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:VVAHNNQLRQON7WYOIU2LDZDAKY","short_pith_number":"pith:VVAHNNQL","schema_version":"1.0","canonical_sha256":"ad4076b60b8c1cdfdb0e4534b1e460562ca5c312a940b5d4011fc0c387b221a0","source":{"kind":"arxiv","id":"2303.02331","version":2},"attestation_state":"computed","paper":{"title":"Training-Free Acceleration of ViTs with Delayed Spatial Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Arash Fayyazi, Jung Hwan Heo, Massoud Pedram, Seyedarmin Azizi","submitted_at":"2023-03-04T05:34:25Z","abstract_excerpt":"Token merging has emerged as a new paradigm that can accelerate the inference of Vision Transformers (ViTs) without any retraining or fine-tuning. To push the frontier of training-free acceleration in ViTs, we improve token merging by adding the perspectives of 1) activation outliers and 2) hierarchical representations. Through a careful analysis of the attention behavior in ViTs, we characterize a delayed onset of the convergent attention phenomenon, which makes token merging undesirable in the bottom blocks of ViTs. Moreover, we augment token merging with a hierarchical processing scheme 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":"2303.02331","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-03-04T05:34:25Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"f359bad7a566a593f045dbc7387caeded4a0743e2c5e97df13b062d85572ecf4","abstract_canon_sha256":"82fc2c61423eb6e315c4e132dc8243911e38450e19b8ce2b300effbf276b308b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:38:32.566228Z","signature_b64":"5w0y/kvlukIxK1E5h2gVpYlcTgDacjAsw3rpTp8KsF35cVTy27Mr6JYmAdZuqAUPNBlEFIRfDQsSdJnbEuUWCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ad4076b60b8c1cdfdb0e4534b1e460562ca5c312a940b5d4011fc0c387b221a0","last_reissued_at":"2026-07-05T08:38:32.565735Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:38:32.565735Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training-Free Acceleration of ViTs with Delayed Spatial Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Arash Fayyazi, Jung Hwan Heo, Massoud Pedram, Seyedarmin Azizi","submitted_at":"2023-03-04T05:34:25Z","abstract_excerpt":"Token merging has emerged as a new paradigm that can accelerate the inference of Vision Transformers (ViTs) without any retraining or fine-tuning. To push the frontier of training-free acceleration in ViTs, we improve token merging by adding the perspectives of 1) activation outliers and 2) hierarchical representations. Through a careful analysis of the attention behavior in ViTs, we characterize a delayed onset of the convergent attention phenomenon, which makes token merging undesirable in the bottom blocks of ViTs. Moreover, we augment token merging with a hierarchical processing scheme to "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.02331","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/2303.02331/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":"2303.02331","created_at":"2026-07-05T08:38:32.565792+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.02331v2","created_at":"2026-07-05T08:38:32.565792+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.02331","created_at":"2026-07-05T08:38:32.565792+00:00"},{"alias_kind":"pith_short_12","alias_value":"VVAHNNQLRQON","created_at":"2026-07-05T08:38:32.565792+00:00"},{"alias_kind":"pith_short_16","alias_value":"VVAHNNQLRQON7WYO","created_at":"2026-07-05T08:38:32.565792+00:00"},{"alias_kind":"pith_short_8","alias_value":"VVAHNNQL","created_at":"2026-07-05T08:38:32.565792+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.10569","citing_title":"Learning to Merge Tokens via Decoupled Embedding for Efficient Vision Transformers","ref_index":12,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY","json":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY.json","graph_json":"https://pith.science/api/pith-number/VVAHNNQLRQON7WYOIU2LDZDAKY/graph.json","events_json":"https://pith.science/api/pith-number/VVAHNNQLRQON7WYOIU2LDZDAKY/events.json","paper":"https://pith.science/paper/VVAHNNQL"},"agent_actions":{"view_html":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY","download_json":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY.json","view_paper":"https://pith.science/paper/VVAHNNQL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.02331&json=true","fetch_graph":"https://pith.science/api/pith-number/VVAHNNQLRQON7WYOIU2LDZDAKY/graph.json","fetch_events":"https://pith.science/api/pith-number/VVAHNNQLRQON7WYOIU2LDZDAKY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY/action/storage_attestation","attest_author":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY/action/author_attestation","sign_citation":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY/action/citation_signature","submit_replication":"https://pith.science/pith/VVAHNNQLRQON7WYOIU2LDZDAKY/action/replication_record"}},"created_at":"2026-07-05T08:38:32.565792+00:00","updated_at":"2026-07-05T08:38:32.565792+00:00"}