{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:R7Q4DEHVPPODLIOCUU6K4PHYCL","short_pith_number":"pith:R7Q4DEHV","canonical_record":{"source":{"id":"2604.24278","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2026-04-27T10:11:08Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"7e0b972031496ec25541affdda8331fd5105533229ed92e5fbd4cfee8efa39d8","abstract_canon_sha256":"5c0668636b798d7392c9049aa6ef9d8e981d058c013b9edf4683536ebfd15a90"},"schema_version":"1.0"},"canonical_sha256":"8fe1c190f57bdc35a1c2a53cae3cf812cbc9e0fa4976ea1c07887458a69388f2","source":{"kind":"arxiv","id":"2604.24278","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.24278","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.24278v3","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.24278","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"pith_short_12","alias_value":"R7Q4DEHVPPOD","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"pith_short_16","alias_value":"R7Q4DEHVPPODLIOC","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"pith_short_8","alias_value":"R7Q4DEHV","created_at":"2026-06-05T01:14:39Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:R7Q4DEHVPPODLIOCUU6K4PHYCL","target":"record","payload":{"canonical_record":{"source":{"id":"2604.24278","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2026-04-27T10:11:08Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"7e0b972031496ec25541affdda8331fd5105533229ed92e5fbd4cfee8efa39d8","abstract_canon_sha256":"5c0668636b798d7392c9049aa6ef9d8e981d058c013b9edf4683536ebfd15a90"},"schema_version":"1.0"},"canonical_sha256":"8fe1c190f57bdc35a1c2a53cae3cf812cbc9e0fa4976ea1c07887458a69388f2","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:14:39.618846Z","signature_b64":"d23OXY7d30AS0ADo5Nnytva345vTITKjqdO3AAGJ+m1UukyT8xwjce4CWMhcbnzM8sULisEs4e/AkF8lQtJJCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8fe1c190f57bdc35a1c2a53cae3cf812cbc9e0fa4976ea1c07887458a69388f2","last_reissued_at":"2026-06-05T01:14:39.618049Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:14:39.618049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.24278","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-05T01:14:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7l1sAXXGWyu3YZDoHn38YR643XSfF+qlCJzlvFOLp3/Qv3knlPqfleN0wZ0fZ2bWMO/SnpPJ9kM1spRbQzBLBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T19:11:21.286897Z"},"content_sha256":"004c9139d76e9f263776d35cbac8eb527f9be3d8a163ebb855c3dbb7fb39c103","schema_version":"1.0","event_id":"sha256:004c9139d76e9f263776d35cbac8eb527f9be3d8a163ebb855c3dbb7fb39c103"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:R7Q4DEHVPPODLIOCUU6K4PHYCL","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"RAS: a Reliability Oriented Metric for Automatic Speech Recognition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ASR models can be trained to abstain from uncertain segments using a human-calibrated reliability metric that improves trustworthiness without losing much accuracy.","cross_cats":["cs.AI"],"primary_cat":"cs.SD","authors_text":"Bohan Li, Hankun Wang, Jing Peng, Kai Yu, Wenbin Huang, Xie Chen, Yiwei Guo, Yuhang Qiu","submitted_at":"2026-04-27T10:11:08Z","abstract_excerpt":"Automatic speech recognition systems often produce confident yet incorrect transcriptions under noisy or ambiguous conditions, which can be misleading for both users and downstream applications. Standard evaluation based on Word Error Rate focuses solely on accuracy and fails to capture transcription reliability. We introduce an abstention-aware transcription framework that enables ASR models to explicitly abstain from uncertain segments. To evaluate reliability under abstention, we propose RAS, a reliability-oriented metric that balances transcription informativeness and error aversion, with "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We introduce an abstention-aware transcription framework that enables ASR models to explicitly abstain from uncertain segments. [...] We propose RAS, a reliability-oriented metric that balances transcription informativeness and error aversion, with its trade-off parameter calibrated by human preference. [...] Our experiments demonstrate substantial improvements in transcription reliability while maintaining competitive accuracy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That human preferences used to calibrate the RAS trade-off parameter are stable, representative, and produce a metric whose optimization via reinforcement learning generalizes beyond the training conditions without introducing new failure modes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RAS is a reliability metric for ASR that trades off informativeness against error aversion via human-calibrated abstention, trained with supervised bootstrapping plus reinforcement learning to improve reliability while keeping accuracy competitive.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ASR models can be trained to abstain from uncertain segments using a human-calibrated reliability metric that improves trustworthiness without losing much accuracy.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6ca2ed908e9b3e02831ce86b74f9bcdb85608dfd3b80c0ab7269d79a1970eadf"},"source":{"id":"2604.24278","kind":"arxiv","version":3},"verdict":{"id":"2e82b9cd-8b6f-42ea-adaf-45a4e36c36cd","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T17:48:58.078600Z","strongest_claim":"We introduce an abstention-aware transcription framework that enables ASR models to explicitly abstain from uncertain segments. [...] We propose RAS, a reliability-oriented metric that balances transcription informativeness and error aversion, with its trade-off parameter calibrated by human preference. [...] Our experiments demonstrate substantial improvements in transcription reliability while maintaining competitive accuracy.","one_line_summary":"RAS is a reliability metric for ASR that trades off informativeness against error aversion via human-calibrated abstention, trained with supervised bootstrapping plus reinforcement learning to improve reliability while keeping accuracy competitive.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That human preferences used to calibrate the RAS trade-off parameter are stable, representative, and produce a metric whose optimization via reinforcement learning generalizes beyond the training conditions without introducing new failure modes.","pith_extraction_headline":"ASR models can be trained to abstain from uncertain segments using a human-calibrated reliability metric that improves trustworthiness without losing much accuracy."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.24278/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T07:35:03.879932Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:16:39.653547Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"37f50fd6eea891c80359b53abab1b877b70ebd73e6b1ba4531c8fb5685fd9825"},"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":"2e82b9cd-8b6f-42ea-adaf-45a4e36c36cd"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-05T01:14:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"NR8VgilJH4Xz35g++ff8qFMCcnpWhuLjqB96qF/iG5oL2rltLokqtIFAz63sUF++gcX83stGIeiHhxbcqjVeAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T19:11:21.287733Z"},"content_sha256":"924f50ff27c21530fd82a9caa939bf63e5aca244df7000cd2d9f79a4b471d724","schema_version":"1.0","event_id":"sha256:924f50ff27c21530fd82a9caa939bf63e5aca244df7000cd2d9f79a4b471d724"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL/bundle.json","state_url":"https://pith.science/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL/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":"2026-08-06T19:11:21Z","links":{"resolver":"https://pith.science/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL","bundle":"https://pith.science/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL/bundle.json","state":"https://pith.science/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL/state.json","well_known_bundle":"https://pith.science/.well-known/pith/R7Q4DEHVPPODLIOCUU6K4PHYCL/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:R7Q4DEHVPPODLIOCUU6K4PHYCL","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":"5c0668636b798d7392c9049aa6ef9d8e981d058c013b9edf4683536ebfd15a90","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2026-04-27T10:11:08Z","title_canon_sha256":"7e0b972031496ec25541affdda8331fd5105533229ed92e5fbd4cfee8efa39d8"},"schema_version":"1.0","source":{"id":"2604.24278","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.24278","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.24278v3","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.24278","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"pith_short_12","alias_value":"R7Q4DEHVPPOD","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"pith_short_16","alias_value":"R7Q4DEHVPPODLIOC","created_at":"2026-06-05T01:14:39Z"},{"alias_kind":"pith_short_8","alias_value":"R7Q4DEHV","created_at":"2026-06-05T01:14:39Z"}],"graph_snapshots":[{"event_id":"sha256:924f50ff27c21530fd82a9caa939bf63e5aca244df7000cd2d9f79a4b471d724","target":"graph","created_at":"2026-06-05T01:14:39Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We introduce an abstention-aware transcription framework that enables ASR models to explicitly abstain from uncertain segments. [...] We propose RAS, a reliability-oriented metric that balances transcription informativeness and error aversion, with its trade-off parameter calibrated by human preference. [...] Our experiments demonstrate substantial improvements in transcription reliability while maintaining competitive accuracy."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That human preferences used to calibrate the RAS trade-off parameter are stable, representative, and produce a metric whose optimization via reinforcement learning generalizes beyond the training conditions without introducing new failure modes."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"RAS is a reliability metric for ASR that trades off informativeness against error aversion via human-calibrated abstention, trained with supervised bootstrapping plus reinforcement learning to improve reliability while keeping accuracy competitive."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"ASR models can be trained to abstain from uncertain segments using a human-calibrated reliability metric that improves trustworthiness without losing much accuracy."}],"snapshot_sha256":"6ca2ed908e9b3e02831ce86b74f9bcdb85608dfd3b80c0ab7269d79a1970eadf"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-21T07:35:03.879932Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T22:16:39.653547Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.24278/integrity.json","findings":[],"snapshot_sha256":"37f50fd6eea891c80359b53abab1b877b70ebd73e6b1ba4531c8fb5685fd9825","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Automatic speech recognition systems often produce confident yet incorrect transcriptions under noisy or ambiguous conditions, which can be misleading for both users and downstream applications. Standard evaluation based on Word Error Rate focuses solely on accuracy and fails to capture transcription reliability. We introduce an abstention-aware transcription framework that enables ASR models to explicitly abstain from uncertain segments. To evaluate reliability under abstention, we propose RAS, a reliability-oriented metric that balances transcription informativeness and error aversion, with ","authors_text":"Bohan Li, Hankun Wang, Jing Peng, Kai Yu, Wenbin Huang, Xie Chen, Yiwei Guo, Yuhang Qiu","cross_cats":["cs.AI"],"headline":"ASR models can be trained to abstain from uncertain segments using a human-calibrated reliability metric that improves trustworthiness without losing much accuracy.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2026-04-27T10:11:08Z","title":"RAS: a Reliability Oriented Metric for Automatic Speech Recognition"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.24278","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-07T17:48:58.078600Z","id":"2e82b9cd-8b6f-42ea-adaf-45a4e36c36cd","model_set":{"reader":"grok-4.3"},"one_line_summary":"RAS is a reliability metric for ASR that trades off informativeness against error aversion via human-calibrated abstention, trained with supervised bootstrapping plus reinforcement learning to improve reliability while keeping accuracy competitive.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"ASR models can be trained to abstain from uncertain segments using a human-calibrated reliability metric that improves trustworthiness without losing much accuracy.","strongest_claim":"We introduce an abstention-aware transcription framework that enables ASR models to explicitly abstain from uncertain segments. [...] We propose RAS, a reliability-oriented metric that balances transcription informativeness and error aversion, with its trade-off parameter calibrated by human preference. [...] Our experiments demonstrate substantial improvements in transcription reliability while maintaining competitive accuracy.","weakest_assumption":"That human preferences used to calibrate the RAS trade-off parameter are stable, representative, and produce a metric whose optimization via reinforcement learning generalizes beyond the training conditions without introducing new failure modes."}},"verdict_id":"2e82b9cd-8b6f-42ea-adaf-45a4e36c36cd"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:004c9139d76e9f263776d35cbac8eb527f9be3d8a163ebb855c3dbb7fb39c103","target":"record","created_at":"2026-06-05T01:14:39Z","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":"5c0668636b798d7392c9049aa6ef9d8e981d058c013b9edf4683536ebfd15a90","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2026-04-27T10:11:08Z","title_canon_sha256":"7e0b972031496ec25541affdda8331fd5105533229ed92e5fbd4cfee8efa39d8"},"schema_version":"1.0","source":{"id":"2604.24278","kind":"arxiv","version":3}},"canonical_sha256":"8fe1c190f57bdc35a1c2a53cae3cf812cbc9e0fa4976ea1c07887458a69388f2","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8fe1c190f57bdc35a1c2a53cae3cf812cbc9e0fa4976ea1c07887458a69388f2","first_computed_at":"2026-06-05T01:14:39.618049Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-05T01:14:39.618049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"d23OXY7d30AS0ADo5Nnytva345vTITKjqdO3AAGJ+m1UukyT8xwjce4CWMhcbnzM8sULisEs4e/AkF8lQtJJCQ==","signature_status":"signed_v1","signed_at":"2026-06-05T01:14:39.618846Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.24278","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:004c9139d76e9f263776d35cbac8eb527f9be3d8a163ebb855c3dbb7fb39c103","sha256:924f50ff27c21530fd82a9caa939bf63e5aca244df7000cd2d9f79a4b471d724"],"state_sha256":"85785c4242293b58b1162641ebd3caf1bd4e122e91f2b4fd848f9a51bab93e68"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d/iBiV5icqqj8Nyso+dXZiGSPmkkScxfUppp7NwDbfPiSdPlva8Mgew5PDaEjKMi+7u813hmZ185B/oLKKVxCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T19:11:21.296761Z","bundle_sha256":"c61997a576ce36939a8262a2ec6edefde8b785e5ba74f3e2c5b7c23bbc9be8a1"}}