{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3EO6Q6YMQCIGZAJB7JIIZ4XSZR","short_pith_number":"pith:3EO6Q6YM","schema_version":"1.0","canonical_sha256":"d91de87b0c80906c8121fa508cf2f2cc557e46737b2756f2f01f1ed94948dbb4","source":{"kind":"arxiv","id":"2504.06949","version":2},"attestation_state":"computed","paper":{"title":"Adaptive Computation Pruning for the Forgetting Transformer","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Aaron Courville, Johan Obando-Ceron, Xu Owen He, Zhixuan Lin","submitted_at":"2025-04-09T14:57:55Z","abstract_excerpt":"The recently proposed Forgetting Transformer (FoX) incorporates a forget gate into softmax attention and has shown consistently better or on-par performance compared to the standard RoPE-based Transformer. Notably, many attention heads in FoX tend to forget quickly, causing their output at each timestep to rely primarily on local context. Based on this observation, we propose Adaptive Computation Pruning (ACP) for FoX, a method that dynamically prunes computations involving input-output dependencies that are strongly decayed by the forget gate. In particular, our method performs provably safe "},"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":"2504.06949","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-09T14:57:55Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"44c4a736a32eae0ce84c97f7ef084960f57119fc4d2b9e1633670c6ec67e8f3e","abstract_canon_sha256":"3526ca7b61a13e6da29d436c4c5e15eb93cb213d624a257be2eaa29ffa1fce57"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:52:19.766933Z","signature_b64":"sjObLt3EERg6AQbycNFczv/ye+n4Ky9aciuOgeu2QG9YbeYelrH7nIs5W4tGOJS6LfZdnpeFLygzM6T7wNICCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d91de87b0c80906c8121fa508cf2f2cc557e46737b2756f2f01f1ed94948dbb4","last_reissued_at":"2026-07-05T11:52:19.766430Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:52:19.766430Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adaptive Computation Pruning for the Forgetting Transformer","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Aaron Courville, Johan Obando-Ceron, Xu Owen He, Zhixuan Lin","submitted_at":"2025-04-09T14:57:55Z","abstract_excerpt":"The recently proposed Forgetting Transformer (FoX) incorporates a forget gate into softmax attention and has shown consistently better or on-par performance compared to the standard RoPE-based Transformer. Notably, many attention heads in FoX tend to forget quickly, causing their output at each timestep to rely primarily on local context. Based on this observation, we propose Adaptive Computation Pruning (ACP) for FoX, a method that dynamically prunes computations involving input-output dependencies that are strongly decayed by the forget gate. In particular, our method performs provably safe "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.06949","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/2504.06949/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":"2504.06949","created_at":"2026-07-05T11:52:19.766491+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.06949v2","created_at":"2026-07-05T11:52:19.766491+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.06949","created_at":"2026-07-05T11:52:19.766491+00:00"},{"alias_kind":"pith_short_12","alias_value":"3EO6Q6YMQCIG","created_at":"2026-07-05T11:52:19.766491+00:00"},{"alias_kind":"pith_short_16","alias_value":"3EO6Q6YMQCIGZAJB","created_at":"2026-07-05T11:52:19.766491+00:00"},{"alias_kind":"pith_short_8","alias_value":"3EO6Q6YM","created_at":"2026-07-05T11:52:19.766491+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.21056","citing_title":"Radiomics-Guided Vision Transformers for Survival Analysis","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR","json":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR.json","graph_json":"https://pith.science/api/pith-number/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/graph.json","events_json":"https://pith.science/api/pith-number/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/events.json","paper":"https://pith.science/paper/3EO6Q6YM"},"agent_actions":{"view_html":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR","download_json":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR.json","view_paper":"https://pith.science/paper/3EO6Q6YM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.06949&json=true","fetch_graph":"https://pith.science/api/pith-number/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/graph.json","fetch_events":"https://pith.science/api/pith-number/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/action/storage_attestation","attest_author":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/action/author_attestation","sign_citation":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/action/citation_signature","submit_replication":"https://pith.science/pith/3EO6Q6YMQCIGZAJB7JIIZ4XSZR/action/replication_record"}},"created_at":"2026-07-05T11:52:19.766491+00:00","updated_at":"2026-07-05T11:52:19.766491+00:00"}