{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2017:ID3OQ6AQUAAXSSNKRF7EFBQMNX","short_pith_number":"pith:ID3OQ6AQ","canonical_record":{"source":{"id":"1705.04400","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2017-05-11T23:34:42Z","cross_cats_sorted":[],"title_canon_sha256":"cce5a5386241ddd33bafe671b5fe77190540a939149e51d42e740cdbb02a1a8d","abstract_canon_sha256":"bf1c6d62d377469290f3dbab96cda033874814f0835095ea1bf71224c03e2f80"},"schema_version":"1.0"},"canonical_sha256":"40f6e87810a0017949aa897e42860c6de67a9458c7d507b8f17764bba5c0e4a6","source":{"kind":"arxiv","id":"1705.04400","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"1705.04400","created_at":"2026-05-18T00:44:38Z"},{"alias_kind":"arxiv_version","alias_value":"1705.04400v1","created_at":"2026-05-18T00:44:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1705.04400","created_at":"2026-05-18T00:44:38Z"},{"alias_kind":"pith_short_12","alias_value":"ID3OQ6AQUAAX","created_at":"2026-05-18T12:31:21Z"},{"alias_kind":"pith_short_16","alias_value":"ID3OQ6AQUAAXSSNK","created_at":"2026-05-18T12:31:21Z"},{"alias_kind":"pith_short_8","alias_value":"ID3OQ6AQ","created_at":"2026-05-18T12:31:21Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2017:ID3OQ6AQUAAXSSNKRF7EFBQMNX","target":"record","payload":{"canonical_record":{"source":{"id":"1705.04400","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2017-05-11T23:34:42Z","cross_cats_sorted":[],"title_canon_sha256":"cce5a5386241ddd33bafe671b5fe77190540a939149e51d42e740cdbb02a1a8d","abstract_canon_sha256":"bf1c6d62d377469290f3dbab96cda033874814f0835095ea1bf71224c03e2f80"},"schema_version":"1.0"},"canonical_sha256":"40f6e87810a0017949aa897e42860c6de67a9458c7d507b8f17764bba5c0e4a6","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:44:38.218271Z","signature_b64":"aHc1eNUwSEVWaaP+d89DY8WgvgSXUKFJjmz+Du+oN/HLw+DrBNejUtBpIH1XFvUrsjuQ3h/5GCib7zExsUc5DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"40f6e87810a0017949aa897e42860c6de67a9458c7d507b8f17764bba5c0e4a6","last_reissued_at":"2026-05-18T00:44:38.217860Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:44:38.217860Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"1705.04400","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T00:44:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+NTHdUrhMvOHljhjAitHTCtZNdzts536rh/2hZIyRXkhl22FILaYZUapprjpu+i+T9TGChtesEOpDk9gcVo5CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-04T10:15:41.039597Z"},"content_sha256":"7805aff75f50ffa7e47504fa39bd67dfe33a16c33b0e4ecce790a885ceec7b5f","schema_version":"1.0","event_id":"sha256:7805aff75f50ffa7e47504fa39bd67dfe33a16c33b0e4ecce790a885ceec7b5f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2017:ID3OQ6AQUAAXSSNKRF7EFBQMNX","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Reducing Bias in Production Speech Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adam Coates, Ajay Kannan, Anuroop Sriram, Atul Kumar, Christopher Fougner, David Seetapun, Eric Battenberg, Hairong Liu, Heewoo Jun, Jiaji Huang, Markus Kliegl, Rewon Child, Sanjeev Satheesh, Vinay Rao, Yashesh Gaur, Zhenyao Zhu","submitted_at":"2017-05-11T23:34:42Z","abstract_excerpt":"Replacing hand-engineered pipelines with end-to-end deep learning systems has enabled strong results in applications like speech and object recognition. However, the causality and latency constraints of production systems put end-to-end speech models back into the underfitting regime and expose biases in the model that we show cannot be overcome by \"scaling up\", i.e., training bigger models on more data. In this work we systematically identify and address sources of bias, reducing error rates by up to 20% while remaining practical for deployment. We achieve this by utilizing improved neural ar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1705.04400","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":""},"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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T00:44:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YAOTlYVAlwVUp8N9rGiYlxmuGSuj/Gz/ZnmVNxLYIJvNJsyNs8N043SWrVQou3Sdc6USMXUexBx3QE6LDAMTCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-04T10:15:41.039953Z"},"content_sha256":"5d7f10dc37341ac7e4501662dae7648cf4ec3c767f4bd4bd45337c702948afb9","schema_version":"1.0","event_id":"sha256:5d7f10dc37341ac7e4501662dae7648cf4ec3c767f4bd4bd45337c702948afb9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX/bundle.json","state_url":"https://pith.science/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-04T10:15:41Z","links":{"resolver":"https://pith.science/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX","bundle":"https://pith.science/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX/bundle.json","state":"https://pith.science/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ID3OQ6AQUAAXSSNKRF7EFBQMNX/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2017:ID3OQ6AQUAAXSSNKRF7EFBQMNX","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"bf1c6d62d377469290f3dbab96cda033874814f0835095ea1bf71224c03e2f80","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2017-05-11T23:34:42Z","title_canon_sha256":"cce5a5386241ddd33bafe671b5fe77190540a939149e51d42e740cdbb02a1a8d"},"schema_version":"1.0","source":{"id":"1705.04400","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"1705.04400","created_at":"2026-05-18T00:44:38Z"},{"alias_kind":"arxiv_version","alias_value":"1705.04400v1","created_at":"2026-05-18T00:44:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1705.04400","created_at":"2026-05-18T00:44:38Z"},{"alias_kind":"pith_short_12","alias_value":"ID3OQ6AQUAAX","created_at":"2026-05-18T12:31:21Z"},{"alias_kind":"pith_short_16","alias_value":"ID3OQ6AQUAAXSSNK","created_at":"2026-05-18T12:31:21Z"},{"alias_kind":"pith_short_8","alias_value":"ID3OQ6AQ","created_at":"2026-05-18T12:31:21Z"}],"graph_snapshots":[{"event_id":"sha256:5d7f10dc37341ac7e4501662dae7648cf4ec3c767f4bd4bd45337c702948afb9","target":"graph","created_at":"2026-05-18T00:44:38Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Replacing hand-engineered pipelines with end-to-end deep learning systems has enabled strong results in applications like speech and object recognition. However, the causality and latency constraints of production systems put end-to-end speech models back into the underfitting regime and expose biases in the model that we show cannot be overcome by \"scaling up\", i.e., training bigger models on more data. In this work we systematically identify and address sources of bias, reducing error rates by up to 20% while remaining practical for deployment. We achieve this by utilizing improved neural ar","authors_text":"Adam Coates, Ajay Kannan, Anuroop Sriram, Atul Kumar, Christopher Fougner, David Seetapun, Eric Battenberg, Hairong Liu, Heewoo Jun, Jiaji Huang, Markus Kliegl, Rewon Child, Sanjeev Satheesh, Vinay Rao, Yashesh Gaur, Zhenyao Zhu","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2017-05-11T23:34:42Z","title":"Reducing Bias in Production Speech Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1705.04400","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:7805aff75f50ffa7e47504fa39bd67dfe33a16c33b0e4ecce790a885ceec7b5f","target":"record","created_at":"2026-05-18T00:44:38Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"bf1c6d62d377469290f3dbab96cda033874814f0835095ea1bf71224c03e2f80","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2017-05-11T23:34:42Z","title_canon_sha256":"cce5a5386241ddd33bafe671b5fe77190540a939149e51d42e740cdbb02a1a8d"},"schema_version":"1.0","source":{"id":"1705.04400","kind":"arxiv","version":1}},"canonical_sha256":"40f6e87810a0017949aa897e42860c6de67a9458c7d507b8f17764bba5c0e4a6","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"40f6e87810a0017949aa897e42860c6de67a9458c7d507b8f17764bba5c0e4a6","first_computed_at":"2026-05-18T00:44:38.217860Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T00:44:38.217860Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"aHc1eNUwSEVWaaP+d89DY8WgvgSXUKFJjmz+Du+oN/HLw+DrBNejUtBpIH1XFvUrsjuQ3h/5GCib7zExsUc5DA==","signature_status":"signed_v1","signed_at":"2026-05-18T00:44:38.218271Z","signed_message":"canonical_sha256_bytes"},"source_id":"1705.04400","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7805aff75f50ffa7e47504fa39bd67dfe33a16c33b0e4ecce790a885ceec7b5f","sha256:5d7f10dc37341ac7e4501662dae7648cf4ec3c767f4bd4bd45337c702948afb9"],"state_sha256":"7faf32fd8d1dd3e66c0f60b4577cd2bffd2148f54cf1400207e412c26a72b30a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IHuWq1yYHPTp7ZCuwaPQzzFy7U//WDnxs4PZq8nAzc9emHWBjd/OEWuiUq8MUt1sJYZt/0Z/HfqoH2lJ5gBtCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-04T10:15:41.041968Z","bundle_sha256":"ed5fbe47bc882e75fecf17959e267bd30ddaea091d55fc34340a17087fbb43d6"}}