{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:77C5CWBJHFPSZDV6N7AZP733QU","short_pith_number":"pith:77C5CWBJ","schema_version":"1.0","canonical_sha256":"ffc5d15829395f2c8ebe6fc197ff7b851af6618e1e0b92d0a6587cbef8c69c21","source":{"kind":"arxiv","id":"2505.18644","version":1},"attestation_state":"computed","paper":{"title":"Enhancing Generalization of Speech Large Language Models with Multi-Task Behavior Imitation and Speech-Text Interleaving","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.SD"],"primary_cat":"eess.AS","authors_text":"Hui Wang, Jingran Xie, Xiang Li, Xixin Wu, Yang Xiang, Yue Yu, Zhiyong Wu","submitted_at":"2025-05-24T11:09:13Z","abstract_excerpt":"Large language models (LLMs) have shown remarkable generalization across tasks, leading to increased interest in integrating speech with LLMs. These speech LLMs (SLLMs) typically use supervised fine-tuning to align speech with text-based LLMs. However, the lack of annotated speech data across a wide range of tasks hinders alignment efficiency, resulting in poor generalization. To address these issues, we propose a novel multi-task 'behavior imitation' method with speech-text interleaving, called MTBI, which relies solely on paired speech and transcripts. By ensuring the LLM decoder generates e"},"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":"2505.18644","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"eess.AS","submitted_at":"2025-05-24T11:09:13Z","cross_cats_sorted":["cs.CL","cs.SD"],"title_canon_sha256":"09f564fe6db062a244b024291eda6fc0412584a61a17444baec6f8f174a4fbc5","abstract_canon_sha256":"ea49d5362b7f020fdf3159d8b933e4d6cb02ab96082d5652c5980b461184f2a4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:16.976133Z","signature_b64":"/TFjCgF6KTBBh/Zpgn8tfWxzKbrPfCXCfQCuaH6oMrWq1A3hhBdVoURKISimx5A3mQx2NC56l91Gs4AZwcSqCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ffc5d15829395f2c8ebe6fc197ff7b851af6618e1e0b92d0a6587cbef8c69c21","last_reissued_at":"2026-07-05T11:09:16.975632Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:16.975632Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Enhancing Generalization of Speech Large Language Models with Multi-Task Behavior Imitation and Speech-Text Interleaving","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.SD"],"primary_cat":"eess.AS","authors_text":"Hui Wang, Jingran Xie, Xiang Li, Xixin Wu, Yang Xiang, Yue Yu, Zhiyong Wu","submitted_at":"2025-05-24T11:09:13Z","abstract_excerpt":"Large language models (LLMs) have shown remarkable generalization across tasks, leading to increased interest in integrating speech with LLMs. These speech LLMs (SLLMs) typically use supervised fine-tuning to align speech with text-based LLMs. However, the lack of annotated speech data across a wide range of tasks hinders alignment efficiency, resulting in poor generalization. To address these issues, we propose a novel multi-task 'behavior imitation' method with speech-text interleaving, called MTBI, which relies solely on paired speech and transcripts. By ensuring the LLM decoder generates e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18644","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/2505.18644/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":"2505.18644","created_at":"2026-07-05T11:09:16.975701+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.18644v1","created_at":"2026-07-05T11:09:16.975701+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18644","created_at":"2026-07-05T11:09:16.975701+00:00"},{"alias_kind":"pith_short_12","alias_value":"77C5CWBJHFPS","created_at":"2026-07-05T11:09:16.975701+00:00"},{"alias_kind":"pith_short_16","alias_value":"77C5CWBJHFPSZDV6","created_at":"2026-07-05T11:09:16.975701+00:00"},{"alias_kind":"pith_short_8","alias_value":"77C5CWBJ","created_at":"2026-07-05T11:09:16.975701+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01733","citing_title":"Rethinking Speech-LLM Integration for ASR: Effective Joint Speech-Text Training by Interleaving","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2509.14804","citing_title":"Towards Building Speech Large Language Models for Multitask Understanding in Low-Resource Languages","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU","json":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU.json","graph_json":"https://pith.science/api/pith-number/77C5CWBJHFPSZDV6N7AZP733QU/graph.json","events_json":"https://pith.science/api/pith-number/77C5CWBJHFPSZDV6N7AZP733QU/events.json","paper":"https://pith.science/paper/77C5CWBJ"},"agent_actions":{"view_html":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU","download_json":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU.json","view_paper":"https://pith.science/paper/77C5CWBJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.18644&json=true","fetch_graph":"https://pith.science/api/pith-number/77C5CWBJHFPSZDV6N7AZP733QU/graph.json","fetch_events":"https://pith.science/api/pith-number/77C5CWBJHFPSZDV6N7AZP733QU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU/action/storage_attestation","attest_author":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU/action/author_attestation","sign_citation":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU/action/citation_signature","submit_replication":"https://pith.science/pith/77C5CWBJHFPSZDV6N7AZP733QU/action/replication_record"}},"created_at":"2026-07-05T11:09:16.975701+00:00","updated_at":"2026-07-05T11:09:16.975701+00:00"}