{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:GO6ZHSPP5TFS2X4J3UJTV3NRHF","short_pith_number":"pith:GO6ZHSPP","schema_version":"1.0","canonical_sha256":"33bd93c9efeccb2d5f89dd133aedb1394a0ed1a148cc0d63a0a032e22d884445","source":{"kind":"arxiv","id":"2105.14849","version":2},"attestation_state":"computed","paper":{"title":"Why does CTC result in peaky behavior?","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.NE","cs.SD","eess.AS","math.ST","stat.TH"],"primary_cat":"cs.LG","authors_text":"Albert Zeyer, Hermann Ney, Ralf Schl\\\"uter","submitted_at":"2021-05-31T10:03:14Z","abstract_excerpt":"The peaky behavior of CTC models is well known experimentally. However, an understanding about why peaky behavior occurs is missing, and whether this is a good property. We provide a formal analysis of the peaky behavior and gradient descent convergence properties of the CTC loss and related training criteria. Our analysis provides a deep understanding why peaky behavior occurs and when it is suboptimal. On a simple example which should be trivial to learn for any model, we prove that a feed-forward neural network trained with CTC from uniform initialization converges towards peaky behavior wi"},"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":"2105.14849","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2021-05-31T10:03:14Z","cross_cats_sorted":["cs.AI","cs.CL","cs.NE","cs.SD","eess.AS","math.ST","stat.TH"],"title_canon_sha256":"7c552a5f9e4120222c73ea0e927ac17f95e1dd4ac0f51e7b691e03f2c8f35c0b","abstract_canon_sha256":"d8004760caaa0150474f69f7a747fb7cfa190cf4a5a62958f95cc17a7300bef2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:46:14.344716Z","signature_b64":"K/dWcVQjGgg3qcQ42mpFaOF+e8RQF9kXwXpIAd4xpkRhcG+Z0OSDmJvtMwNT0kdfb1LV+cBscFD75+beT+exAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"33bd93c9efeccb2d5f89dd133aedb1394a0ed1a148cc0d63a0a032e22d884445","last_reissued_at":"2026-07-05T02:46:14.344250Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:46:14.344250Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why does CTC result in peaky behavior?","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.NE","cs.SD","eess.AS","math.ST","stat.TH"],"primary_cat":"cs.LG","authors_text":"Albert Zeyer, Hermann Ney, Ralf Schl\\\"uter","submitted_at":"2021-05-31T10:03:14Z","abstract_excerpt":"The peaky behavior of CTC models is well known experimentally. However, an understanding about why peaky behavior occurs is missing, and whether this is a good property. We provide a formal analysis of the peaky behavior and gradient descent convergence properties of the CTC loss and related training criteria. Our analysis provides a deep understanding why peaky behavior occurs and when it is suboptimal. On a simple example which should be trivial to learn for any model, we prove that a feed-forward neural network trained with CTC from uniform initialization converges towards peaky behavior wi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2105.14849","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/2105.14849/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":"2105.14849","created_at":"2026-07-05T02:46:14.344307+00:00"},{"alias_kind":"arxiv_version","alias_value":"2105.14849v2","created_at":"2026-07-05T02:46:14.344307+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2105.14849","created_at":"2026-07-05T02:46:14.344307+00:00"},{"alias_kind":"pith_short_12","alias_value":"GO6ZHSPP5TFS","created_at":"2026-07-05T02:46:14.344307+00:00"},{"alias_kind":"pith_short_16","alias_value":"GO6ZHSPP5TFS2X4J","created_at":"2026-07-05T02:46:14.344307+00:00"},{"alias_kind":"pith_short_8","alias_value":"GO6ZHSPP","created_at":"2026-07-05T02:46:14.344307+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.19048","citing_title":"Conserved Kinematic Representations enable Zero-Shot Decoding in Handwriting BCIs","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22133","citing_title":"Beyond Acoustic Sparsity and Linguistic Bias: A Prompt-Free Paradigm for Mispronunciation Detection and Diagnosis","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08003","citing_title":"Rethinking Entropy Allocation in LLM-based ASR: Understanding the Dynamics between Speech Encoders and LLMs","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF","json":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF.json","graph_json":"https://pith.science/api/pith-number/GO6ZHSPP5TFS2X4J3UJTV3NRHF/graph.json","events_json":"https://pith.science/api/pith-number/GO6ZHSPP5TFS2X4J3UJTV3NRHF/events.json","paper":"https://pith.science/paper/GO6ZHSPP"},"agent_actions":{"view_html":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF","download_json":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF.json","view_paper":"https://pith.science/paper/GO6ZHSPP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2105.14849&json=true","fetch_graph":"https://pith.science/api/pith-number/GO6ZHSPP5TFS2X4J3UJTV3NRHF/graph.json","fetch_events":"https://pith.science/api/pith-number/GO6ZHSPP5TFS2X4J3UJTV3NRHF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF/action/storage_attestation","attest_author":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF/action/author_attestation","sign_citation":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF/action/citation_signature","submit_replication":"https://pith.science/pith/GO6ZHSPP5TFS2X4J3UJTV3NRHF/action/replication_record"}},"created_at":"2026-07-05T02:46:14.344307+00:00","updated_at":"2026-07-05T02:46:14.344307+00:00"}