{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3ICCMY4A677ITGYS6CWOPOFWI7","short_pith_number":"pith:3ICCMY4A","schema_version":"1.0","canonical_sha256":"da04266380f7fe899b12f0ace7b8b647ec05c868dcf31dd0c9ddf9da1ace9986","source":{"kind":"arxiv","id":"2605.23954","version":1},"attestation_state":"computed","paper":{"title":"EchoDistill:Alignment Noisy-to-Clean Self-Distillation for Robust Audio LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.SD"],"primary_cat":"cs.CL","authors_text":"Cai Yuchen, Chunxi Luo, Gongli Xi, Jie Zhang, Jin Wang, Junhao Dong, Kaiwen Luo, Kun Wang, Liang Lin, Qiankun Li, Yuanhe Zhang, Zhenhong Zhou","submitted_at":"2026-05-11T06:30:25Z","abstract_excerpt":"Audio Large Language Models (ALLMs) are highly vulnerable to real-world noise, which often induces severe semantic drift and hallucinations. Existing robustness methods primarily rely on waveform-level acoustic enhancement, answer-level supervision, or the internal suppression of noise representations. To address these issues, we propose echodistill, an alignment-based noisy-to-clean self-distillation framework. Echodistill leverages a frozen clean-audio teacher to provide semantic references for an inference-time noisy-audio student. Specifically, the student samples candidate responses under"},"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":"2605.23954","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-11T06:30:25Z","cross_cats_sorted":["cs.AI","cs.SD"],"title_canon_sha256":"24bf83fccf990c1682ff590b65ef35ee45aa6ee4d0adc5adadf6aa8692b20b5f","abstract_canon_sha256":"b990775778709a0dd4a515210070fd8c6839668c162ed211e4223ce6110d3a44"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T00:02:23.645442Z","signature_b64":"kCAvkMlSYApCYBjIULnY5rLnOsHXZEXLasHiogu2U/hupLxc7b/T4sMLkyMhQHbrxd7T0FEuuYVsD72dgQcYCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"da04266380f7fe899b12f0ace7b8b647ec05c868dcf31dd0c9ddf9da1ace9986","last_reissued_at":"2026-05-26T00:02:23.644731Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T00:02:23.644731Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EchoDistill:Alignment Noisy-to-Clean Self-Distillation for Robust Audio LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.SD"],"primary_cat":"cs.CL","authors_text":"Cai Yuchen, Chunxi Luo, Gongli Xi, Jie Zhang, Jin Wang, Junhao Dong, Kaiwen Luo, Kun Wang, Liang Lin, Qiankun Li, Yuanhe Zhang, Zhenhong Zhou","submitted_at":"2026-05-11T06:30:25Z","abstract_excerpt":"Audio Large Language Models (ALLMs) are highly vulnerable to real-world noise, which often induces severe semantic drift and hallucinations. Existing robustness methods primarily rely on waveform-level acoustic enhancement, answer-level supervision, or the internal suppression of noise representations. To address these issues, we propose echodistill, an alignment-based noisy-to-clean self-distillation framework. Echodistill leverages a frozen clean-audio teacher to provide semantic references for an inference-time noisy-audio student. Specifically, the student samples candidate responses under"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.23954","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/2605.23954/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":"2605.23954","created_at":"2026-05-26T00:02:23.644820+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.23954v1","created_at":"2026-05-26T00:02:23.644820+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.23954","created_at":"2026-05-26T00:02:23.644820+00:00"},{"alias_kind":"pith_short_12","alias_value":"3ICCMY4A677I","created_at":"2026-05-26T00:02:23.644820+00:00"},{"alias_kind":"pith_short_16","alias_value":"3ICCMY4A677ITGYS","created_at":"2026-05-26T00:02:23.644820+00:00"},{"alias_kind":"pith_short_8","alias_value":"3ICCMY4A","created_at":"2026-05-26T00:02:23.644820+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7","json":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7.json","graph_json":"https://pith.science/api/pith-number/3ICCMY4A677ITGYS6CWOPOFWI7/graph.json","events_json":"https://pith.science/api/pith-number/3ICCMY4A677ITGYS6CWOPOFWI7/events.json","paper":"https://pith.science/paper/3ICCMY4A"},"agent_actions":{"view_html":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7","download_json":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7.json","view_paper":"https://pith.science/paper/3ICCMY4A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.23954&json=true","fetch_graph":"https://pith.science/api/pith-number/3ICCMY4A677ITGYS6CWOPOFWI7/graph.json","fetch_events":"https://pith.science/api/pith-number/3ICCMY4A677ITGYS6CWOPOFWI7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7/action/storage_attestation","attest_author":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7/action/author_attestation","sign_citation":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7/action/citation_signature","submit_replication":"https://pith.science/pith/3ICCMY4A677ITGYS6CWOPOFWI7/action/replication_record"}},"created_at":"2026-05-26T00:02:23.644820+00:00","updated_at":"2026-05-26T00:02:23.644820+00:00"}