{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:IGNM4CYO5MDKMKE3WKKHENBT7N","short_pith_number":"pith:IGNM4CYO","schema_version":"1.0","canonical_sha256":"419ace0b0eeb06a6289bb294723433fb40465fe112d2fab7729ecc947b9ad691","source":{"kind":"arxiv","id":"1901.08987","version":2},"attestation_state":"computed","paper":{"title":"Dynamical Isometry and a Mean Field Theory of LSTMs and GRUs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Bo Chang, Dar Gilboa, Ed H. Chi, Greg Yang, Jeffrey Pennington, Minmin Chen, Samuel S. Schoenholz","submitted_at":"2019-01-25T17:05:54Z","abstract_excerpt":"Training recurrent neural networks (RNNs) on long sequence tasks is plagued with difficulties arising from the exponential explosion or vanishing of signals as they propagate forward or backward through the network. Many techniques have been proposed to ameliorate these issues, including various algorithmic and architectural modifications. Two of the most successful RNN architectures, the LSTM and the GRU, do exhibit modest improvements over vanilla RNN cells, but they still suffer from instabilities when trained on very long sequences. In this work, we develop a mean field theory of signal pr"},"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":"1901.08987","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-01-25T17:05:54Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"053cb30ef5d999c9edff256479da58fe89925d70faa63c7d341c3afb4c97dc51","abstract_canon_sha256":"924098a4a9c99e8858de7db89002f1d3cba055222d59c67e06b3747eb5cf09c8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:45:15.029271Z","signature_b64":"g2lBZkM+lwGkVCUGIF+fLYPntjLpt/vClVvnUhg+BDK0Rpe2xX1l2JP7y5EbMb/rvzwipnaeuQzeQGAtNWANAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"419ace0b0eeb06a6289bb294723433fb40465fe112d2fab7729ecc947b9ad691","last_reissued_at":"2026-05-17T23:45:15.028776Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:45:15.028776Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dynamical Isometry and a Mean Field Theory of LSTMs and GRUs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Bo Chang, Dar Gilboa, Ed H. Chi, Greg Yang, Jeffrey Pennington, Minmin Chen, Samuel S. Schoenholz","submitted_at":"2019-01-25T17:05:54Z","abstract_excerpt":"Training recurrent neural networks (RNNs) on long sequence tasks is plagued with difficulties arising from the exponential explosion or vanishing of signals as they propagate forward or backward through the network. Many techniques have been proposed to ameliorate these issues, including various algorithmic and architectural modifications. Two of the most successful RNN architectures, the LSTM and the GRU, do exhibit modest improvements over vanilla RNN cells, but they still suffer from instabilities when trained on very long sequences. In this work, we develop a mean field theory of signal pr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1901.08987","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":""},"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":"1901.08987","created_at":"2026-05-17T23:45:15.028857+00:00"},{"alias_kind":"arxiv_version","alias_value":"1901.08987v2","created_at":"2026-05-17T23:45:15.028857+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1901.08987","created_at":"2026-05-17T23:45:15.028857+00:00"},{"alias_kind":"pith_short_12","alias_value":"IGNM4CYO5MDK","created_at":"2026-05-18T12:33:18.533446+00:00"},{"alias_kind":"pith_short_16","alias_value":"IGNM4CYO5MDKMKE3","created_at":"2026-05-18T12:33:18.533446+00:00"},{"alias_kind":"pith_short_8","alias_value":"IGNM4CYO","created_at":"2026-05-18T12:33:18.533446+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.12121","citing_title":"Time-Scale Coupling Between States and Parameters in Recurrent Neural Networks","ref_index":22,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05113","citing_title":"How Long Does Infinite Width Last? Signal Propagation in Long-Range Linear Recurrences","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N","json":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N.json","graph_json":"https://pith.science/api/pith-number/IGNM4CYO5MDKMKE3WKKHENBT7N/graph.json","events_json":"https://pith.science/api/pith-number/IGNM4CYO5MDKMKE3WKKHENBT7N/events.json","paper":"https://pith.science/paper/IGNM4CYO"},"agent_actions":{"view_html":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N","download_json":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N.json","view_paper":"https://pith.science/paper/IGNM4CYO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1901.08987&json=true","fetch_graph":"https://pith.science/api/pith-number/IGNM4CYO5MDKMKE3WKKHENBT7N/graph.json","fetch_events":"https://pith.science/api/pith-number/IGNM4CYO5MDKMKE3WKKHENBT7N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N/action/storage_attestation","attest_author":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N/action/author_attestation","sign_citation":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N/action/citation_signature","submit_replication":"https://pith.science/pith/IGNM4CYO5MDKMKE3WKKHENBT7N/action/replication_record"}},"created_at":"2026-05-17T23:45:15.028857+00:00","updated_at":"2026-05-17T23:45:15.028857+00:00"}