{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:COR55WETYG5KRCNM2WMEA6HH4S","short_pith_number":"pith:COR55WET","schema_version":"1.0","canonical_sha256":"13a3ded893c1baa889acd5984078e7e48f2d760058b206024fb4407684c566d1","source":{"kind":"arxiv","id":"2505.19670","version":1},"attestation_state":"computed","paper":{"title":"Reshaping Representation Space to Balance the Safety and Over-rejection in Large Audio Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.MM","cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Ehsan Shareghi, Gholamreza Haffari, Hao Yang, Lizhen Qu","submitted_at":"2025-05-26T08:25:25Z","abstract_excerpt":"Large Audio Language Models (LALMs) have extended the capabilities of Large Language Models (LLMs) by enabling audio-based human interactions. However, recent research has revealed that LALMs remain vulnerable to harmful queries due to insufficient safety-alignment. Despite advances in defence measures for text and vision LLMs, effective safety-alignment strategies and audio-safety dataset specifically targeting LALMs are notably absent. Meanwhile defence measures based on Supervised Fine-tuning (SFT) struggle to address safety improvement while avoiding over-rejection issues, significantly co"},"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":"2505.19670","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-26T08:25:25Z","cross_cats_sorted":["cs.MM","cs.SD","eess.AS"],"title_canon_sha256":"b188cd103359d13b734f44a07a12fec20dd737cb1e5efab9bee4209d95112d79","abstract_canon_sha256":"2d38fdaf377df5a17ddf339de38db1e68713dee9aa7ab30d28d24ab822f4ef59"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:34.261245Z","signature_b64":"eDukDpQZzeFCO1eIjIRu9aX1SmYVUS+lH6eTzc5efHH4kAdLK3TQ702/y2xWIGV7V3eBhpk45InLhSxJ5XGlBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"13a3ded893c1baa889acd5984078e7e48f2d760058b206024fb4407684c566d1","last_reissued_at":"2026-07-05T11:09:34.260768Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:34.260768Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reshaping Representation Space to Balance the Safety and Over-rejection in Large Audio Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.MM","cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Ehsan Shareghi, Gholamreza Haffari, Hao Yang, Lizhen Qu","submitted_at":"2025-05-26T08:25:25Z","abstract_excerpt":"Large Audio Language Models (LALMs) have extended the capabilities of Large Language Models (LLMs) by enabling audio-based human interactions. However, recent research has revealed that LALMs remain vulnerable to harmful queries due to insufficient safety-alignment. Despite advances in defence measures for text and vision LLMs, effective safety-alignment strategies and audio-safety dataset specifically targeting LALMs are notably absent. Meanwhile defence measures based on Supervised Fine-tuning (SFT) struggle to address safety improvement while avoiding over-rejection issues, significantly co"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.19670","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/2505.19670/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":"2505.19670","created_at":"2026-07-05T11:09:34.260820+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.19670v1","created_at":"2026-07-05T11:09:34.260820+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.19670","created_at":"2026-07-05T11:09:34.260820+00:00"},{"alias_kind":"pith_short_12","alias_value":"COR55WETYG5K","created_at":"2026-07-05T11:09:34.260820+00:00"},{"alias_kind":"pith_short_16","alias_value":"COR55WETYG5KRCNM","created_at":"2026-07-05T11:09:34.260820+00:00"},{"alias_kind":"pith_short_8","alias_value":"COR55WET","created_at":"2026-07-05T11:09:34.260820+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21147","citing_title":"AOR-Bench: Do Large Audio Language Models Over-Refuse Pseudo-Harmful Queries?","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20266","citing_title":"A Survey of Large Audio Language Models: Generalization, Trustworthiness, and Outlook","ref_index":173,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S","json":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S.json","graph_json":"https://pith.science/api/pith-number/COR55WETYG5KRCNM2WMEA6HH4S/graph.json","events_json":"https://pith.science/api/pith-number/COR55WETYG5KRCNM2WMEA6HH4S/events.json","paper":"https://pith.science/paper/COR55WET"},"agent_actions":{"view_html":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S","download_json":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S.json","view_paper":"https://pith.science/paper/COR55WET","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.19670&json=true","fetch_graph":"https://pith.science/api/pith-number/COR55WETYG5KRCNM2WMEA6HH4S/graph.json","fetch_events":"https://pith.science/api/pith-number/COR55WETYG5KRCNM2WMEA6HH4S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S/action/storage_attestation","attest_author":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S/action/author_attestation","sign_citation":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S/action/citation_signature","submit_replication":"https://pith.science/pith/COR55WETYG5KRCNM2WMEA6HH4S/action/replication_record"}},"created_at":"2026-07-05T11:09:34.260820+00:00","updated_at":"2026-07-05T11:09:34.260820+00:00"}