{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3APZTNOU6JEGQV76FR45TZR64O","short_pith_number":"pith:3APZTNOU","schema_version":"1.0","canonical_sha256":"d81f99b5d4f2486857fe2c79d9e63ee39657e61b85661318dd2149272d2c1fca","source":{"kind":"arxiv","id":"2505.17852","version":1},"attestation_state":"computed","paper":{"title":"Scaling Recurrent Neural Networks to a Billion Parameters with Zero-Order Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Francois Chaubard, Mykel Kochenderfer","submitted_at":"2025-05-23T13:04:06Z","abstract_excerpt":"During inference, Recurrent Neural Networks (RNNs) scale constant in both FLOPs and GPU memory with increasing context length, as they compress all prior tokens into a fixed-size memory. In contrast, transformers scale linearly in FLOPs and, at best, linearly in memory during generation, since they must attend to all previous tokens explicitly. Despite this inference-time advantage, training large RNNs on long contexts remains impractical because standard optimization methods depend on Backpropagation Through Time (BPTT). BPTT requires retention of all intermediate activations during the forwa"},"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":"2505.17852","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-23T13:04:06Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c9a2fdaba493dcce4b87ee7c840eb1a7e1000a1ce609310782a29ba4f085ea1d","abstract_canon_sha256":"71a0aee2709c4c5410138886b173b867e187228125e15f03353e29dd20ec1a4f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:32.887283Z","signature_b64":"N1PP7rMtaFV4EqEOHDMbSM4MYxUWCw9I859i+0L6fBvk91KOeStjn7+Nj+eEfH0CeEesVSeN6phULGLyqcC6Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d81f99b5d4f2486857fe2c79d9e63ee39657e61b85661318dd2149272d2c1fca","last_reissued_at":"2026-07-05T11:08:32.886808Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:32.886808Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Recurrent Neural Networks to a Billion Parameters with Zero-Order Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Francois Chaubard, Mykel Kochenderfer","submitted_at":"2025-05-23T13:04:06Z","abstract_excerpt":"During inference, Recurrent Neural Networks (RNNs) scale constant in both FLOPs and GPU memory with increasing context length, as they compress all prior tokens into a fixed-size memory. In contrast, transformers scale linearly in FLOPs and, at best, linearly in memory during generation, since they must attend to all previous tokens explicitly. Despite this inference-time advantage, training large RNNs on long contexts remains impractical because standard optimization methods depend on Backpropagation Through Time (BPTT). BPTT requires retention of all intermediate activations during the forwa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.17852","kind":"arxiv","version":1},"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/2505.17852/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":"2505.17852","created_at":"2026-07-05T11:08:32.886866+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.17852v1","created_at":"2026-07-05T11:08:32.886866+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.17852","created_at":"2026-07-05T11:08:32.886866+00:00"},{"alias_kind":"pith_short_12","alias_value":"3APZTNOU6JEG","created_at":"2026-07-05T11:08:32.886866+00:00"},{"alias_kind":"pith_short_16","alias_value":"3APZTNOU6JEGQV76","created_at":"2026-07-05T11:08:32.886866+00:00"},{"alias_kind":"pith_short_8","alias_value":"3APZTNOU","created_at":"2026-07-05T11:08:32.886866+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.15113","citing_title":"Low-rank surrogate modeling and stochastic zero-order optimization for training of neural networks with black-box layers","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O","json":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O.json","graph_json":"https://pith.science/api/pith-number/3APZTNOU6JEGQV76FR45TZR64O/graph.json","events_json":"https://pith.science/api/pith-number/3APZTNOU6JEGQV76FR45TZR64O/events.json","paper":"https://pith.science/paper/3APZTNOU"},"agent_actions":{"view_html":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O","download_json":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O.json","view_paper":"https://pith.science/paper/3APZTNOU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.17852&json=true","fetch_graph":"https://pith.science/api/pith-number/3APZTNOU6JEGQV76FR45TZR64O/graph.json","fetch_events":"https://pith.science/api/pith-number/3APZTNOU6JEGQV76FR45TZR64O/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O/action/storage_attestation","attest_author":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O/action/author_attestation","sign_citation":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O/action/citation_signature","submit_replication":"https://pith.science/pith/3APZTNOU6JEGQV76FR45TZR64O/action/replication_record"}},"created_at":"2026-07-05T11:08:32.886866+00:00","updated_at":"2026-07-05T11:08:32.886866+00:00"}