{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:T3ZKZ4XH2ZM27CKFQXLMWDLLYR","short_pith_number":"pith:T3ZKZ4XH","schema_version":"1.0","canonical_sha256":"9ef2acf2e7d659af894585d6cb0d6bc45578f474057b213f978aa32cd2207783","source":{"kind":"arxiv","id":"2107.00910","version":3},"attestation_state":"computed","paper":{"title":"Learned Token Pruning for Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amir Gholami, David Thorsley, Joseph Hassoun, Kurt Keutzer, Sehoon Kim, Sheng Shen, Woosuk Kwon","submitted_at":"2021-07-02T09:00:13Z","abstract_excerpt":"Deploying transformer models in practice is challenging due to their inference cost, which scales quadratically with input sequence length. To address this, we present a novel Learned Token Pruning (LTP) method which adaptively removes unimportant tokens as an input sequence passes through transformer layers. In particular, LTP prunes tokens with an attention score below a threshold value which is learned for each layer during training. Our threshold-based method allows the length of the pruned sequence to vary adaptively based on the input sequence, and avoids algorithmically expensive operat"},"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":"2107.00910","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-07-02T09:00:13Z","cross_cats_sorted":[],"title_canon_sha256":"c15c0364bab3f605ac1d88f10e02844cc83994f7bbcff000178137c8084beb0f","abstract_canon_sha256":"d10c26887b45844fcf6ffba41a9a67c2d59299738aab1d10296c150d169e84a3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:28:40.679447Z","signature_b64":"SXYXz2YtlL/2Yxl2zhf/8utyPfqnvlIGy94tjYgqctfw4sVpi0FqAHCMpPc7LwCYQquTfAWRCG1B4IK8u7lHAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ef2acf2e7d659af894585d6cb0d6bc45578f474057b213f978aa32cd2207783","last_reissued_at":"2026-07-05T04:28:40.678906Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:28:40.678906Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learned Token Pruning for Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amir Gholami, David Thorsley, Joseph Hassoun, Kurt Keutzer, Sehoon Kim, Sheng Shen, Woosuk Kwon","submitted_at":"2021-07-02T09:00:13Z","abstract_excerpt":"Deploying transformer models in practice is challenging due to their inference cost, which scales quadratically with input sequence length. To address this, we present a novel Learned Token Pruning (LTP) method which adaptively removes unimportant tokens as an input sequence passes through transformer layers. In particular, LTP prunes tokens with an attention score below a threshold value which is learned for each layer during training. Our threshold-based method allows the length of the pruned sequence to vary adaptively based on the input sequence, and avoids algorithmically expensive operat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2107.00910","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/2107.00910/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":"2107.00910","created_at":"2026-07-05T04:28:40.678973+00:00"},{"alias_kind":"arxiv_version","alias_value":"2107.00910v3","created_at":"2026-07-05T04:28:40.678973+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2107.00910","created_at":"2026-07-05T04:28:40.678973+00:00"},{"alias_kind":"pith_short_12","alias_value":"T3ZKZ4XH2ZM2","created_at":"2026-07-05T04:28:40.678973+00:00"},{"alias_kind":"pith_short_16","alias_value":"T3ZKZ4XH2ZM27CKF","created_at":"2026-07-05T04:28:40.678973+00:00"},{"alias_kind":"pith_short_8","alias_value":"T3ZKZ4XH","created_at":"2026-07-05T04:28:40.678973+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.29480","citing_title":"DTM-Codec: Dynamic Token Masking for VFR Speech Coding with Efficient Boundary Selection","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2210.09461","citing_title":"Token Merging: Your ViT But Faster","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR","json":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR.json","graph_json":"https://pith.science/api/pith-number/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/graph.json","events_json":"https://pith.science/api/pith-number/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/events.json","paper":"https://pith.science/paper/T3ZKZ4XH"},"agent_actions":{"view_html":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR","download_json":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR.json","view_paper":"https://pith.science/paper/T3ZKZ4XH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2107.00910&json=true","fetch_graph":"https://pith.science/api/pith-number/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/graph.json","fetch_events":"https://pith.science/api/pith-number/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/action/storage_attestation","attest_author":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/action/author_attestation","sign_citation":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/action/citation_signature","submit_replication":"https://pith.science/pith/T3ZKZ4XH2ZM27CKFQXLMWDLLYR/action/replication_record"}},"created_at":"2026-07-05T04:28:40.678973+00:00","updated_at":"2026-07-05T04:28:40.678973+00:00"}