{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ICAEJSIFTW6SBWZ7MABIPQWVVL","short_pith_number":"pith:ICAEJSIF","schema_version":"1.0","canonical_sha256":"408044c9059dbd20db3f600287c2d5aaca3e08a5abe75229a76a16db2b9d8a3a","source":{"kind":"arxiv","id":"2508.15882","version":1},"attestation_state":"computed","paper":{"title":"Beyond Transcription: Mechanistic Interpretability in ASR","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG","eess.AS"],"primary_cat":"cs.SD","authors_text":"Asaf Buchnick, Aviv Navon, Aviv Shamsian, Ethan Fetaya, Gill Hetz, Hilit Segev, Joseph Keshet, Neta Glazer, Yael Segal-Feldman","submitted_at":"2025-08-21T15:42:53Z","abstract_excerpt":"Interpretability methods have recently gained significant attention, particularly in the context of large language models, enabling insights into linguistic representations, error detection, and model behaviors such as hallucinations and repetitions. However, these techniques remain underexplored in automatic speech recognition (ASR), despite their potential to advance both the performance and interpretability of ASR systems. In this work, we adapt and systematically apply established interpretability methods such as logit lens, linear probing, and activation patching, to examine how acoustic "},"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":"2508.15882","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SD","submitted_at":"2025-08-21T15:42:53Z","cross_cats_sorted":["cs.CL","cs.LG","eess.AS"],"title_canon_sha256":"99a3b322e4bc73650260b0fa9bcfea531d18d955809626f81747aaf94b3df5dc","abstract_canon_sha256":"6814650fd9b42cc5ef51783ee49cb431295c3cebf18d27f1198ef4abb0e7a2da"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:22.388850Z","signature_b64":"So/MV/A6MxCUcU5+AOc1mE7soPkk/VfxCvYUXlDs3Y0fdkj2X4UbaTkD3nhAU0PKfhdJIAIZNdLifPjY13ppBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"408044c9059dbd20db3f600287c2d5aaca3e08a5abe75229a76a16db2b9d8a3a","last_reissued_at":"2026-07-05T11:57:22.388355Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:22.388355Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Transcription: Mechanistic Interpretability in ASR","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG","eess.AS"],"primary_cat":"cs.SD","authors_text":"Asaf Buchnick, Aviv Navon, Aviv Shamsian, Ethan Fetaya, Gill Hetz, Hilit Segev, Joseph Keshet, Neta Glazer, Yael Segal-Feldman","submitted_at":"2025-08-21T15:42:53Z","abstract_excerpt":"Interpretability methods have recently gained significant attention, particularly in the context of large language models, enabling insights into linguistic representations, error detection, and model behaviors such as hallucinations and repetitions. However, these techniques remain underexplored in automatic speech recognition (ASR), despite their potential to advance both the performance and interpretability of ASR systems. In this work, we adapt and systematically apply established interpretability methods such as logit lens, linear probing, and activation patching, to examine how acoustic "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.15882","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.15882/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":"2508.15882","created_at":"2026-07-05T11:57:22.388435+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.15882v1","created_at":"2026-07-05T11:57:22.388435+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.15882","created_at":"2026-07-05T11:57:22.388435+00:00"},{"alias_kind":"pith_short_12","alias_value":"ICAEJSIFTW6S","created_at":"2026-07-05T11:57:22.388435+00:00"},{"alias_kind":"pith_short_16","alias_value":"ICAEJSIFTW6SBWZ7","created_at":"2026-07-05T11:57:22.388435+00:00"},{"alias_kind":"pith_short_8","alias_value":"ICAEJSIF","created_at":"2026-07-05T11:57:22.388435+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23060","citing_title":"From Text Metrics to Model Internals: A Study of Whisper ASR Hallucination Detection","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23048","citing_title":"HALAS: A Human-Annotated Dataset of Hallucinations of Modern ASR Systems","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22473","citing_title":"Interleaved Speech Language Models Latently Work In Text","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL","json":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL.json","graph_json":"https://pith.science/api/pith-number/ICAEJSIFTW6SBWZ7MABIPQWVVL/graph.json","events_json":"https://pith.science/api/pith-number/ICAEJSIFTW6SBWZ7MABIPQWVVL/events.json","paper":"https://pith.science/paper/ICAEJSIF"},"agent_actions":{"view_html":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL","download_json":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL.json","view_paper":"https://pith.science/paper/ICAEJSIF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.15882&json=true","fetch_graph":"https://pith.science/api/pith-number/ICAEJSIFTW6SBWZ7MABIPQWVVL/graph.json","fetch_events":"https://pith.science/api/pith-number/ICAEJSIFTW6SBWZ7MABIPQWVVL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL/action/storage_attestation","attest_author":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL/action/author_attestation","sign_citation":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL/action/citation_signature","submit_replication":"https://pith.science/pith/ICAEJSIFTW6SBWZ7MABIPQWVVL/action/replication_record"}},"created_at":"2026-07-05T11:57:22.388435+00:00","updated_at":"2026-07-05T11:57:22.388435+00:00"}