{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:V33TZU25JQRQZ4225OE67AMTTU","short_pith_number":"pith:V33TZU25","schema_version":"1.0","canonical_sha256":"aef73cd35d4c230cf35aeb89ef81939d1356c62b13477fe81d90d3f0e53a26a2","source":{"kind":"arxiv","id":"2606.23052","version":1},"attestation_state":"computed","paper":{"title":"CAAD: Contrastive Audio-Aware Distillation for Efficient Speech Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"eess.AS","authors_text":"Chun-Wei Chen, Hung-yi Lee, Ke-Han Lu, Tzu-Quan Lin, Wei-Ping Huang","submitted_at":"2026-06-22T09:03:45Z","abstract_excerpt":"Speech Language Models achieve reasoning capabilities, but are often hindered by massive parameter counts and a tendency to prioritize linguistic priors over acoustic features. While contrastive decoding enhances grounding by contrasting audio-aware and text-only logits, it increases inference latency. We propose Contrastive Audio-Aware Distillation (CAAD), a framework that internalizes the teacher's contrastive reasoning into the student model's weights. To overcome the high computational training overhead in the dual-path token-by-token contrastive distillation process, we introduce a synchr"},"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":"2606.23052","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"eess.AS","submitted_at":"2026-06-22T09:03:45Z","cross_cats_sorted":[],"title_canon_sha256":"a94c680b6753e087b686b46a42ae9e53b83d1eb6946c06b0e2132c6065d30a95","abstract_canon_sha256":"49de5381c9688e16542a8e1274979a13c7a90148268d2261ae1e9c65edbd7c60"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T03:14:08.324915Z","signature_b64":"4/A80FC2U7LD42xjNOGsQnyMh2Gbyz01Fj72xNFR+MKWPPdJK2TmLu2kZyJo4pjVYZRmNkCgySLyKOazCCXsBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aef73cd35d4c230cf35aeb89ef81939d1356c62b13477fe81d90d3f0e53a26a2","last_reissued_at":"2026-06-23T03:14:08.324506Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T03:14:08.324506Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CAAD: Contrastive Audio-Aware Distillation for Efficient Speech Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"eess.AS","authors_text":"Chun-Wei Chen, Hung-yi Lee, Ke-Han Lu, Tzu-Quan Lin, Wei-Ping Huang","submitted_at":"2026-06-22T09:03:45Z","abstract_excerpt":"Speech Language Models achieve reasoning capabilities, but are often hindered by massive parameter counts and a tendency to prioritize linguistic priors over acoustic features. While contrastive decoding enhances grounding by contrasting audio-aware and text-only logits, it increases inference latency. We propose Contrastive Audio-Aware Distillation (CAAD), a framework that internalizes the teacher's contrastive reasoning into the student model's weights. To overcome the high computational training overhead in the dual-path token-by-token contrastive distillation process, we introduce a synchr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.23052","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/2606.23052/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":"2606.23052","created_at":"2026-06-23T03:14:08.324562+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.23052v1","created_at":"2026-06-23T03:14:08.324562+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.23052","created_at":"2026-06-23T03:14:08.324562+00:00"},{"alias_kind":"pith_short_12","alias_value":"V33TZU25JQRQ","created_at":"2026-06-23T03:14:08.324562+00:00"},{"alias_kind":"pith_short_16","alias_value":"V33TZU25JQRQZ422","created_at":"2026-06-23T03:14:08.324562+00:00"},{"alias_kind":"pith_short_8","alias_value":"V33TZU25","created_at":"2026-06-23T03:14:08.324562+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.23052","citing_title":"CAAD: Contrastive Audio-Aware Distillation for Efficient Speech Language Models","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU","json":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU.json","graph_json":"https://pith.science/api/pith-number/V33TZU25JQRQZ4225OE67AMTTU/graph.json","events_json":"https://pith.science/api/pith-number/V33TZU25JQRQZ4225OE67AMTTU/events.json","paper":"https://pith.science/paper/V33TZU25"},"agent_actions":{"view_html":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU","download_json":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU.json","view_paper":"https://pith.science/paper/V33TZU25","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.23052&json=true","fetch_graph":"https://pith.science/api/pith-number/V33TZU25JQRQZ4225OE67AMTTU/graph.json","fetch_events":"https://pith.science/api/pith-number/V33TZU25JQRQZ4225OE67AMTTU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU/action/storage_attestation","attest_author":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU/action/author_attestation","sign_citation":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU/action/citation_signature","submit_replication":"https://pith.science/pith/V33TZU25JQRQZ4225OE67AMTTU/action/replication_record"}},"created_at":"2026-06-23T03:14:08.324562+00:00","updated_at":"2026-06-23T03:14:08.324562+00:00"}