{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RTQMQ6SAOZ7UFW6DWPJA5XT54Z","short_pith_number":"pith:RTQMQ6SA","schema_version":"1.0","canonical_sha256":"8ce0c87a40767f42dbc3b3d20ede7de67bc686f414c742df74b58721ab1e2f81","source":{"kind":"arxiv","id":"2602.23958","version":2},"attestation_state":"computed","paper":{"title":"An Empirical Analysis of Task-Induced Encoder Bias in Fr\\'echet Audio Distance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SD"],"primary_cat":"eess.AS","authors_text":"Wonwoo Jeong","submitted_at":"2026-02-27T12:01:16Z","abstract_excerpt":"Fr\\'echet Audio Distance (FAD) is the de facto standard for evaluating text-to-audio generation, yet its scores depend on the underlying encoder's embedding space. An encoder's training task dictates which acoustic features are preserved or discarded, causing FAD to inherit systematic task-induced biases. We decompose evaluation into Recall, Precision, and Alignment (split into semantic and structural dimensions), using log-scale normalization for fair cross-encoder comparison. Controlled experiments on six encoders across two datasets reveal a four-axis trade-off: reconstruction-based AudioMA"},"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":"2602.23958","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"eess.AS","submitted_at":"2026-02-27T12:01:16Z","cross_cats_sorted":["cs.SD"],"title_canon_sha256":"4ba0b7b7f7c0fba69c70109f8d4d07e024a837930a9d63fe267fa1617f872def","abstract_canon_sha256":"e453e381f18fd2f5ad204893d7e1fe0723af43f903461988b646385fdeb7755d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:05:15.192727Z","signature_b64":"Bj8jP3TIq7JnEykHIo4JyZvVVFxDO14i8d4nD7KdGVvCoBmuurczLhqJeFqoqKa7JhRAImpFKzc5uJU8uOc+CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8ce0c87a40767f42dbc3b3d20ede7de67bc686f414c742df74b58721ab1e2f81","last_reissued_at":"2026-06-09T01:05:15.192237Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:05:15.192237Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"An Empirical Analysis of Task-Induced Encoder Bias in Fr\\'echet Audio Distance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SD"],"primary_cat":"eess.AS","authors_text":"Wonwoo Jeong","submitted_at":"2026-02-27T12:01:16Z","abstract_excerpt":"Fr\\'echet Audio Distance (FAD) is the de facto standard for evaluating text-to-audio generation, yet its scores depend on the underlying encoder's embedding space. An encoder's training task dictates which acoustic features are preserved or discarded, causing FAD to inherit systematic task-induced biases. We decompose evaluation into Recall, Precision, and Alignment (split into semantic and structural dimensions), using log-scale normalization for fair cross-encoder comparison. Controlled experiments on six encoders across two datasets reveal a four-axis trade-off: reconstruction-based AudioMA"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.23958","kind":"arxiv","version":2},"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/2602.23958/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":"2602.23958","created_at":"2026-06-09T01:05:15.192297+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.23958v2","created_at":"2026-06-09T01:05:15.192297+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.23958","created_at":"2026-06-09T01:05:15.192297+00:00"},{"alias_kind":"pith_short_12","alias_value":"RTQMQ6SAOZ7U","created_at":"2026-06-09T01:05:15.192297+00:00"},{"alias_kind":"pith_short_16","alias_value":"RTQMQ6SAOZ7UFW6D","created_at":"2026-06-09T01:05:15.192297+00:00"},{"alias_kind":"pith_short_8","alias_value":"RTQMQ6SA","created_at":"2026-06-09T01:05:15.192297+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.05554","citing_title":"Optimal Transport Audio Distance with Learned Riemannian Ground Metrics","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2604.10542","citing_title":"VidAudio-Bench: Benchmarking V2A and VT2A Generation across Four Audio Categories","ref_index":26,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z","json":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z.json","graph_json":"https://pith.science/api/pith-number/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/graph.json","events_json":"https://pith.science/api/pith-number/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/events.json","paper":"https://pith.science/paper/RTQMQ6SA"},"agent_actions":{"view_html":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z","download_json":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z.json","view_paper":"https://pith.science/paper/RTQMQ6SA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.23958&json=true","fetch_graph":"https://pith.science/api/pith-number/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/graph.json","fetch_events":"https://pith.science/api/pith-number/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/action/storage_attestation","attest_author":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/action/author_attestation","sign_citation":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/action/citation_signature","submit_replication":"https://pith.science/pith/RTQMQ6SAOZ7UFW6DWPJA5XT54Z/action/replication_record"}},"created_at":"2026-06-09T01:05:15.192297+00:00","updated_at":"2026-06-09T01:05:15.192297+00:00"}