{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OXGLJCVNJKVYDCVQINPNXIX6LU","short_pith_number":"pith:OXGLJCVN","schema_version":"1.0","canonical_sha256":"75ccb48aad4aab818ab0435edba2fe5d000ccd2261d813b1024d9629e9d0cdae","source":{"kind":"arxiv","id":"2502.07827","version":3},"attestation_state":"computed","paper":{"title":"Implicit Language Models are RNNs: Balancing Parallelization and Expressivity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Babak Rahmani, Fabian Falck, Heiner Kremer, Hitesh Ballani, Jannes Gladrow, Mark Sch\\\"one","submitted_at":"2025-02-10T19:59:31Z","abstract_excerpt":"State-space models (SSMs) and transformers dominate the language modeling landscape. However, they are constrained to a lower computational complexity than classical recurrent neural networks (RNNs), limiting their expressivity. In contrast, RNNs lack parallelization during training, raising fundamental questions about the trade off between parallelization and expressivity. We propose implicit SSMs, which iterate a transformation until convergence to a fixed point. Theoretically, we show that implicit SSMs implement the non-linear state-transitions of RNNs. Empirically, we find that only appro"},"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":"2502.07827","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-10T19:59:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0207e18b919cdc9a1710a3bef5679635a6b68fa12a361df3f990b54ab26b1021","abstract_canon_sha256":"263fac45250631d2a3e88e23c1265f5dcbe91c9e5fb910e24bbdc0a8c6e004a9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:09.031291Z","signature_b64":"Za8K4X5OqS8nmfsnEsNcL5ZbJP6PASX2JRqfq0Rk63nxYqfSArOFlRT/CMyKi+OEdS6iqFiJlSWz81sWhe41Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"75ccb48aad4aab818ab0435edba2fe5d000ccd2261d813b1024d9629e9d0cdae","last_reissued_at":"2026-07-05T11:20:09.030785Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:09.030785Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Implicit Language Models are RNNs: Balancing Parallelization and Expressivity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Babak Rahmani, Fabian Falck, Heiner Kremer, Hitesh Ballani, Jannes Gladrow, Mark Sch\\\"one","submitted_at":"2025-02-10T19:59:31Z","abstract_excerpt":"State-space models (SSMs) and transformers dominate the language modeling landscape. However, they are constrained to a lower computational complexity than classical recurrent neural networks (RNNs), limiting their expressivity. In contrast, RNNs lack parallelization during training, raising fundamental questions about the trade off between parallelization and expressivity. We propose implicit SSMs, which iterate a transformation until convergence to a fixed point. Theoretically, we show that implicit SSMs implement the non-linear state-transitions of RNNs. Empirically, we find that only appro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.07827","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/2502.07827/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":"2502.07827","created_at":"2026-07-05T11:20:09.030846+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.07827v3","created_at":"2026-07-05T11:20:09.030846+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.07827","created_at":"2026-07-05T11:20:09.030846+00:00"},{"alias_kind":"pith_short_12","alias_value":"OXGLJCVNJKVY","created_at":"2026-07-05T11:20:09.030846+00:00"},{"alias_kind":"pith_short_16","alias_value":"OXGLJCVNJKVYDCVQ","created_at":"2026-07-05T11:20:09.030846+00:00"},{"alias_kind":"pith_short_8","alias_value":"OXGLJCVN","created_at":"2026-07-05T11:20:09.030846+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.15871","citing_title":"Agentic Discovery of Neural Architectures: AIRA-Compose and AIRA-Design","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU","json":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU.json","graph_json":"https://pith.science/api/pith-number/OXGLJCVNJKVYDCVQINPNXIX6LU/graph.json","events_json":"https://pith.science/api/pith-number/OXGLJCVNJKVYDCVQINPNXIX6LU/events.json","paper":"https://pith.science/paper/OXGLJCVN"},"agent_actions":{"view_html":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU","download_json":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU.json","view_paper":"https://pith.science/paper/OXGLJCVN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.07827&json=true","fetch_graph":"https://pith.science/api/pith-number/OXGLJCVNJKVYDCVQINPNXIX6LU/graph.json","fetch_events":"https://pith.science/api/pith-number/OXGLJCVNJKVYDCVQINPNXIX6LU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU/action/storage_attestation","attest_author":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU/action/author_attestation","sign_citation":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU/action/citation_signature","submit_replication":"https://pith.science/pith/OXGLJCVNJKVYDCVQINPNXIX6LU/action/replication_record"}},"created_at":"2026-07-05T11:20:09.030846+00:00","updated_at":"2026-07-05T11:20:09.030846+00:00"}