{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:JW6LBMYUWAY7ERXW7NOCTSVDZZ","short_pith_number":"pith:JW6LBMYU","schema_version":"1.0","canonical_sha256":"4dbcb0b314b031f246f6fb5c29caa3ce7afe221d318812c0b73fe5f8929a2acc","source":{"kind":"arxiv","id":"2311.10770","version":2},"attestation_state":"computed","paper":{"title":"Exponentially Faster Language Modelling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.NE"],"primary_cat":"cs.CL","authors_text":"Peter Belcak, Roger Wattenhofer","submitted_at":"2023-11-15T18:42:50Z","abstract_excerpt":"Language models only really need to use an exponential fraction of their neurons for individual inferences. As proof, we present UltraFastBERT, a BERT variant that uses 0.3% of its neurons during inference while performing on par with similar BERT models. UltraFastBERT selectively engages just 12 out of 4095 neurons for each layer inference. This is achieved by replacing feedforward networks with fast feedforward networks (FFFs). While no truly efficient implementation currently exists to unlock the full acceleration potential of conditional neural execution, we provide high-level CPU code ach"},"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":"2311.10770","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-15T18:42:50Z","cross_cats_sorted":["cs.AI","cs.LG","cs.NE"],"title_canon_sha256":"b7fbb354f5f60dcf6748831c6f194c080f43705fd03a9c3a31c708c6e0348d9f","abstract_canon_sha256":"110cdbf93365d6f069ea9765f07408e21ca85f5ecaf73dbd65ce69271ba25e3b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:14:59.162389Z","signature_b64":"YwejZcM1bvjEB2y8qYHXIF+5naenaDGYnVAdBVX3KGCo7GcaeVlvOeaLexUCdf32u37BHrnU7KghKU0e4rcmDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4dbcb0b314b031f246f6fb5c29caa3ce7afe221d318812c0b73fe5f8929a2acc","last_reissued_at":"2026-07-05T07:14:59.161888Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:14:59.161888Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Exponentially Faster Language Modelling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.NE"],"primary_cat":"cs.CL","authors_text":"Peter Belcak, Roger Wattenhofer","submitted_at":"2023-11-15T18:42:50Z","abstract_excerpt":"Language models only really need to use an exponential fraction of their neurons for individual inferences. As proof, we present UltraFastBERT, a BERT variant that uses 0.3% of its neurons during inference while performing on par with similar BERT models. UltraFastBERT selectively engages just 12 out of 4095 neurons for each layer inference. This is achieved by replacing feedforward networks with fast feedforward networks (FFFs). While no truly efficient implementation currently exists to unlock the full acceleration potential of conditional neural execution, we provide high-level CPU code ach"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.10770","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/2311.10770/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":"2311.10770","created_at":"2026-07-05T07:14:59.161941+00:00"},{"alias_kind":"arxiv_version","alias_value":"2311.10770v2","created_at":"2026-07-05T07:14:59.161941+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.10770","created_at":"2026-07-05T07:14:59.161941+00:00"},{"alias_kind":"pith_short_12","alias_value":"JW6LBMYUWAY7","created_at":"2026-07-05T07:14:59.161941+00:00"},{"alias_kind":"pith_short_16","alias_value":"JW6LBMYUWAY7ERXW","created_at":"2026-07-05T07:14:59.161941+00:00"},{"alias_kind":"pith_short_8","alias_value":"JW6LBMYU","created_at":"2026-07-05T07:14:59.161941+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.30516","citing_title":"HASTE: A Framework for Training-Free, Dynamic, and Steerable Compression of Pre-Trained Convolutional Neural Networks","ref_index":65,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ","json":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ.json","graph_json":"https://pith.science/api/pith-number/JW6LBMYUWAY7ERXW7NOCTSVDZZ/graph.json","events_json":"https://pith.science/api/pith-number/JW6LBMYUWAY7ERXW7NOCTSVDZZ/events.json","paper":"https://pith.science/paper/JW6LBMYU"},"agent_actions":{"view_html":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ","download_json":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ.json","view_paper":"https://pith.science/paper/JW6LBMYU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2311.10770&json=true","fetch_graph":"https://pith.science/api/pith-number/JW6LBMYUWAY7ERXW7NOCTSVDZZ/graph.json","fetch_events":"https://pith.science/api/pith-number/JW6LBMYUWAY7ERXW7NOCTSVDZZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ/action/storage_attestation","attest_author":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ/action/author_attestation","sign_citation":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ/action/citation_signature","submit_replication":"https://pith.science/pith/JW6LBMYUWAY7ERXW7NOCTSVDZZ/action/replication_record"}},"created_at":"2026-07-05T07:14:59.161941+00:00","updated_at":"2026-07-05T07:14:59.161941+00:00"}