{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:P53EMQOYJ4SWOADEQ7BABL4JXN","short_pith_number":"pith:P53EMQOY","schema_version":"1.0","canonical_sha256":"7f764641d84f2567006487c200af89bb7216c0dc1b078dbe49fa7b3c6e95fd6a","source":{"kind":"arxiv","id":"2410.02678","version":1},"attestation_state":"computed","paper":{"title":"Distilling an End-to-End Voice Assistant Without Instruction Training Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Diyi Yang, Ella Li, Michael Ryan, Weiyan Shi, William Held, Yanzhe Zhang","submitted_at":"2024-10-03T17:04:48Z","abstract_excerpt":"Voice assistants, such as Siri and Google Assistant, typically model audio and text separately, resulting in lost speech information and increased complexity. Recent efforts to address this with end-to-end Speech Large Language Models (LLMs) trained with supervised finetuning (SFT)\n  have led to models ``forgetting\" capabilities from text-only LLMs. Our work proposes an alternative paradigm for training Speech LLMs without instruction data, using the response of a text-only LLM to transcripts as self-supervision. Importantly, this process can be performed without annotated responses. We show t"},"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":"2410.02678","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-03T17:04:48Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"50ea97663cb391393e9a3a952d9a857eee4466119fae3abb7a6e51d8fef00bf2","abstract_canon_sha256":"d17108efcd1357b4f3b6972aacdb54846ef5decc0a3fcb8f137c719491a8abdd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:15:24.067303Z","signature_b64":"/Iuk1Xlyb+fLNcp8EIm6uX/iz3jzWK2p8EPaXQLha1IrjPUsmHOYIRJIOzDiz/A/O6cx/VJ9TvfqN8e0CkGTBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7f764641d84f2567006487c200af89bb7216c0dc1b078dbe49fa7b3c6e95fd6a","last_reissued_at":"2026-07-05T09:15:24.066868Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:15:24.066868Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Distilling an End-to-End Voice Assistant Without Instruction Training Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Diyi Yang, Ella Li, Michael Ryan, Weiyan Shi, William Held, Yanzhe Zhang","submitted_at":"2024-10-03T17:04:48Z","abstract_excerpt":"Voice assistants, such as Siri and Google Assistant, typically model audio and text separately, resulting in lost speech information and increased complexity. Recent efforts to address this with end-to-end Speech Large Language Models (LLMs) trained with supervised finetuning (SFT)\n  have led to models ``forgetting\" capabilities from text-only LLMs. Our work proposes an alternative paradigm for training Speech LLMs without instruction data, using the response of a text-only LLM to transcripts as self-supervision. Importantly, this process can be performed without annotated responses. We show t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.02678","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/2410.02678/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":"2410.02678","created_at":"2026-07-05T09:15:24.066928+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.02678v1","created_at":"2026-07-05T09:15:24.066928+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.02678","created_at":"2026-07-05T09:15:24.066928+00:00"},{"alias_kind":"pith_short_12","alias_value":"P53EMQOYJ4SW","created_at":"2026-07-05T09:15:24.066928+00:00"},{"alias_kind":"pith_short_16","alias_value":"P53EMQOYJ4SWOADE","created_at":"2026-07-05T09:15:24.066928+00:00"},{"alias_kind":"pith_short_8","alias_value":"P53EMQOY","created_at":"2026-07-05T09:15:24.066928+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.19858","citing_title":"Benchmarking Gaslighting Attacks Against Speech Large Language Models","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2504.08528","citing_title":"On The Landscape of Spoken Language Models: A Comprehensive Survey","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2410.17196","citing_title":"VoiceBench: Benchmarking LLM-Based Voice Assistants","ref_index":74,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN","json":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN.json","graph_json":"https://pith.science/api/pith-number/P53EMQOYJ4SWOADEQ7BABL4JXN/graph.json","events_json":"https://pith.science/api/pith-number/P53EMQOYJ4SWOADEQ7BABL4JXN/events.json","paper":"https://pith.science/paper/P53EMQOY"},"agent_actions":{"view_html":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN","download_json":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN.json","view_paper":"https://pith.science/paper/P53EMQOY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.02678&json=true","fetch_graph":"https://pith.science/api/pith-number/P53EMQOYJ4SWOADEQ7BABL4JXN/graph.json","fetch_events":"https://pith.science/api/pith-number/P53EMQOYJ4SWOADEQ7BABL4JXN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN/action/storage_attestation","attest_author":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN/action/author_attestation","sign_citation":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN/action/citation_signature","submit_replication":"https://pith.science/pith/P53EMQOYJ4SWOADEQ7BABL4JXN/action/replication_record"}},"created_at":"2026-07-05T09:15:24.066928+00:00","updated_at":"2026-07-05T09:15:24.066928+00:00"}