{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:7MHLXIJN57DLRHLRQLEOJKR62N","short_pith_number":"pith:7MHLXIJN","canonical_record":{"source":{"id":"2604.21928","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T17:59:47Z","cross_cats_sorted":[],"title_canon_sha256":"4ee570597049557e2cae96c44baa57d38b90793d65d543068780c09968dbc8ee","abstract_canon_sha256":"3ac8902258213c474683680a2538adbdab43fa048145918b227caf3e03a98b99"},"schema_version":"1.0"},"canonical_sha256":"fb0ebba12defc6b89d7182c8e4aa3ed34a7fbbb6d46a1cab35c54d0c7613bc68","source":{"kind":"arxiv","id":"2604.21928","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.21928","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"arxiv_version","alias_value":"2604.21928v3","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.21928","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"pith_short_12","alias_value":"7MHLXIJN57DL","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"pith_short_16","alias_value":"7MHLXIJN57DLRHLR","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"pith_short_8","alias_value":"7MHLXIJN","created_at":"2026-06-01T01:02:40Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:7MHLXIJN57DLRHLRQLEOJKR62N","target":"record","payload":{"canonical_record":{"source":{"id":"2604.21928","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T17:59:47Z","cross_cats_sorted":[],"title_canon_sha256":"4ee570597049557e2cae96c44baa57d38b90793d65d543068780c09968dbc8ee","abstract_canon_sha256":"3ac8902258213c474683680a2538adbdab43fa048145918b227caf3e03a98b99"},"schema_version":"1.0"},"canonical_sha256":"fb0ebba12defc6b89d7182c8e4aa3ed34a7fbbb6d46a1cab35c54d0c7613bc68","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:02:40.322181Z","signature_b64":"atQBGvFL6NMToEbbQNB5W8+YU5fXUKNssCFLOiSQfKj51qpRd6MPTm3kbsS0vu4TRYeo/dkr18IUzajIYqFjCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fb0ebba12defc6b89d7182c8e4aa3ed34a7fbbb6d46a1cab35c54d0c7613bc68","last_reissued_at":"2026-06-01T01:02:40.321190Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:02:40.321190Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.21928","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":"2026-06-01T01:02:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7tlMljB1w+vMWTRBbzleGQAxhNSNqWF+NVmLnbuwZmMddDzU2JpsxUa7d7+6CcBrH/H/Rq7KhArrz+Y4SIOGCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-11T20:55:31.346626Z"},"content_sha256":"aec197da2b5488e8de234b94b66448073abc434697935b455407bd65bd36c5a9","schema_version":"1.0","event_id":"sha256:aec197da2b5488e8de234b94b66448073abc434697935b455407bd65bd36c5a9"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:7MHLXIJN57DLRHLRQLEOJKR62N","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Evaluation of Automatic Speech Recognition Using Generative Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Large language models select the better automatic speech recognition hypothesis with 92-94 percent agreement to humans, compared to 63 percent for word error rate.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Driss Khalil, Jane Wottawa, Mickael Rouvier, Petr Motlicek, Richard Dufour, Sergio Burdisso, Shashi Kumar, Shiran Liu, Thibault Ba\\~neras-Roux","submitted_at":"2026-04-23T17:59:47Z","abstract_excerpt":"Automatic Speech Recognition (ASR) is traditionally evaluated using Word Error Rate (WER), a metric that is insensitive to meaning. Embedding-based semantic metrics are better correlated with human perception, but decoder-based Large Language Models (LLMs) remain underexplored for this task. This paper evaluates their relevance through three approaches: (1) selecting the best hypothesis between two candidates, (2) computing semantic distance using generative embeddings, and (3) qualitative classification of errors. On the HATS dataset, the best LLMs achieve 92--94\\% agreement with human annota"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On the HATS dataset, the best LLMs achieve 92--94% agreement with human annotators for hypothesis selection, compared to 63% for WER, also outperforming semantic metrics.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That agreement with human annotators on pairwise hypothesis selection is a sufficient and unbiased proxy for semantic correctness of ASR outputs across domains.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Decoder-based LLMs reach 92-94% agreement with humans when choosing correct ASR hypotheses, beating WER at 63% and outperforming encoder-based semantic metrics.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Large language models select the better automatic speech recognition hypothesis with 92-94 percent agreement to humans, compared to 63 percent for word error rate.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"25fff00d9d056bb6738cc39c9e4061b1231863c24a02270835bad2c6b5b2e54e"},"source":{"id":"2604.21928","kind":"arxiv","version":3},"verdict":{"id":"608d6511-729b-48e8-9179-f9f16b35d079","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-09T21:20:34.343252Z","strongest_claim":"On the HATS dataset, the best LLMs achieve 92--94% agreement with human annotators for hypothesis selection, compared to 63% for WER, also outperforming semantic metrics.","one_line_summary":"Decoder-based LLMs reach 92-94% agreement with humans when choosing correct ASR hypotheses, beating WER at 63% and outperforming encoder-based semantic metrics.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That agreement with human annotators on pairwise hypothesis selection is a sufficient and unbiased proxy for semantic correctness of ASR outputs across domains.","pith_extraction_headline":"Large language models select the better automatic speech recognition hypothesis with 92-94 percent agreement to humans, compared to 63 percent for word error rate."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.21928/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T11:38:38.827905Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T00:27:58.518760Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"d92ed3e6aa5ee4e212b27872d541389c1d22d154fcaa3863ba6bbef3026dcdd0"},"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":"608d6511-729b-48e8-9179-f9f16b35d079"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-01T01:02:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vqpeaBnk8mrfjNmZpQGWA2IqGW3v958BpoDhJQRmrxhe50RO1W9Nhr4uULsS9PQNHr4RN5A/FAYIvo/QMRNfCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-11T20:55:31.347634Z"},"content_sha256":"a1c13098949c8cd853a06ca6ab2df519483b7f80393d9d72a444287bd14c42be","schema_version":"1.0","event_id":"sha256:a1c13098949c8cd853a06ca6ab2df519483b7f80393d9d72a444287bd14c42be"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/7MHLXIJN57DLRHLRQLEOJKR62N/bundle.json","state_url":"https://pith.science/pith/7MHLXIJN57DLRHLRQLEOJKR62N/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/7MHLXIJN57DLRHLRQLEOJKR62N/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":"2026-06-11T20:55:31Z","links":{"resolver":"https://pith.science/pith/7MHLXIJN57DLRHLRQLEOJKR62N","bundle":"https://pith.science/pith/7MHLXIJN57DLRHLRQLEOJKR62N/bundle.json","state":"https://pith.science/pith/7MHLXIJN57DLRHLRQLEOJKR62N/state.json","well_known_bundle":"https://pith.science/.well-known/pith/7MHLXIJN57DLRHLRQLEOJKR62N/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:7MHLXIJN57DLRHLRQLEOJKR62N","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":"3ac8902258213c474683680a2538adbdab43fa048145918b227caf3e03a98b99","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T17:59:47Z","title_canon_sha256":"4ee570597049557e2cae96c44baa57d38b90793d65d543068780c09968dbc8ee"},"schema_version":"1.0","source":{"id":"2604.21928","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.21928","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"arxiv_version","alias_value":"2604.21928v3","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.21928","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"pith_short_12","alias_value":"7MHLXIJN57DL","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"pith_short_16","alias_value":"7MHLXIJN57DLRHLR","created_at":"2026-06-01T01:02:40Z"},{"alias_kind":"pith_short_8","alias_value":"7MHLXIJN","created_at":"2026-06-01T01:02:40Z"}],"graph_snapshots":[{"event_id":"sha256:a1c13098949c8cd853a06ca6ab2df519483b7f80393d9d72a444287bd14c42be","target":"graph","created_at":"2026-06-01T01:02:40Z","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":"On the HATS dataset, the best LLMs achieve 92--94% agreement with human annotators for hypothesis selection, compared to 63% for WER, also outperforming semantic metrics."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That agreement with human annotators on pairwise hypothesis selection is a sufficient and unbiased proxy for semantic correctness of ASR outputs across domains."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Decoder-based LLMs reach 92-94% agreement with humans when choosing correct ASR hypotheses, beating WER at 63% and outperforming encoder-based semantic metrics."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Large language models select the better automatic speech recognition hypothesis with 92-94 percent agreement to humans, compared to 63 percent for word error rate."}],"snapshot_sha256":"25fff00d9d056bb6738cc39c9e4061b1231863c24a02270835bad2c6b5b2e54e"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-21T11:38:38.827905Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-20T00:27:58.518760Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.21928/integrity.json","findings":[],"snapshot_sha256":"d92ed3e6aa5ee4e212b27872d541389c1d22d154fcaa3863ba6bbef3026dcdd0","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Automatic Speech Recognition (ASR) is traditionally evaluated using Word Error Rate (WER), a metric that is insensitive to meaning. Embedding-based semantic metrics are better correlated with human perception, but decoder-based Large Language Models (LLMs) remain underexplored for this task. This paper evaluates their relevance through three approaches: (1) selecting the best hypothesis between two candidates, (2) computing semantic distance using generative embeddings, and (3) qualitative classification of errors. On the HATS dataset, the best LLMs achieve 92--94\\% agreement with human annota","authors_text":"Driss Khalil, Jane Wottawa, Mickael Rouvier, Petr Motlicek, Richard Dufour, Sergio Burdisso, Shashi Kumar, Shiran Liu, Thibault Ba\\~neras-Roux","cross_cats":[],"headline":"Large language models select the better automatic speech recognition hypothesis with 92-94 percent agreement to humans, compared to 63 percent for word error rate.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T17:59:47Z","title":"Evaluation of Automatic Speech Recognition Using Generative Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.21928","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-09T21:20:34.343252Z","id":"608d6511-729b-48e8-9179-f9f16b35d079","model_set":{"reader":"grok-4.3"},"one_line_summary":"Decoder-based LLMs reach 92-94% agreement with humans when choosing correct ASR hypotheses, beating WER at 63% and outperforming encoder-based semantic metrics.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Large language models select the better automatic speech recognition hypothesis with 92-94 percent agreement to humans, compared to 63 percent for word error rate.","strongest_claim":"On the HATS dataset, the best LLMs achieve 92--94% agreement with human annotators for hypothesis selection, compared to 63% for WER, also outperforming semantic metrics.","weakest_assumption":"That agreement with human annotators on pairwise hypothesis selection is a sufficient and unbiased proxy for semantic correctness of ASR outputs across domains."}},"verdict_id":"608d6511-729b-48e8-9179-f9f16b35d079"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:aec197da2b5488e8de234b94b66448073abc434697935b455407bd65bd36c5a9","target":"record","created_at":"2026-06-01T01:02:40Z","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":"3ac8902258213c474683680a2538adbdab43fa048145918b227caf3e03a98b99","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T17:59:47Z","title_canon_sha256":"4ee570597049557e2cae96c44baa57d38b90793d65d543068780c09968dbc8ee"},"schema_version":"1.0","source":{"id":"2604.21928","kind":"arxiv","version":3}},"canonical_sha256":"fb0ebba12defc6b89d7182c8e4aa3ed34a7fbbb6d46a1cab35c54d0c7613bc68","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"fb0ebba12defc6b89d7182c8e4aa3ed34a7fbbb6d46a1cab35c54d0c7613bc68","first_computed_at":"2026-06-01T01:02:40.321190Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-01T01:02:40.321190Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"atQBGvFL6NMToEbbQNB5W8+YU5fXUKNssCFLOiSQfKj51qpRd6MPTm3kbsS0vu4TRYeo/dkr18IUzajIYqFjCg==","signature_status":"signed_v1","signed_at":"2026-06-01T01:02:40.322181Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.21928","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:aec197da2b5488e8de234b94b66448073abc434697935b455407bd65bd36c5a9","sha256:a1c13098949c8cd853a06ca6ab2df519483b7f80393d9d72a444287bd14c42be"],"state_sha256":"d21abe72ce02287d4ecc14904eb783fa823dec00142f4f197d9e7892ebd3647e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"o4zi2zSpQjr3Nxs1pejimrNHsfw0v/jxZFuVxY+tKjH1c25qjWABnIPa8Ed/rzJI7clFlyuIPJMaUUvqR2MpCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-11T20:55:31.352372Z","bundle_sha256":"0f03bf3e330d72336193ee2b42acc9d3186c4367a5b1e1d74fed9e7066b29082"}}