{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:D7FKJYJKJVZGVYKBLF5KGT5G77","short_pith_number":"pith:D7FKJYJK","schema_version":"1.0","canonical_sha256":"1fcaa4e12a4d726ae141597aa34fa6ffc2e88baa8410f0e5a5dc797674fe4ae6","source":{"kind":"arxiv","id":"2510.14737","version":3},"attestation_state":"computed","paper":{"title":"Free-Grained Hierarchical Visual Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Hierarchical image recognition can learn consistent predictions from labels at any taxonomy level using mixed-granularity supervision.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Seulki Park, Stella X. Yu, Zilin Wang","submitted_at":"2025-10-16T14:35:18Z","abstract_excerpt":"Hierarchical image recognition seeks to predict class labels along a semantic taxonomy, from broad categories to specific ones, typically under the tidy assumption that every training image is fully annotated along its taxonomy path. Reality is messier: A distant bird may be labeled only bird, while a clear close-up may justify bald eagle. We introduce free-grain training, where labels may appear at any level of the taxonomy and models must learn consistent hierarchical predictions from incomplete, mixed-granularity supervision. We build benchmark datasets with varying label granularity and sh"},"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":true},"canonical_record":{"source":{"id":"2510.14737","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-10-16T14:35:18Z","cross_cats_sorted":[],"title_canon_sha256":"c85882320849c4d4692746168094e519189c8a6ad98d90be1bed9d6d7b174313","abstract_canon_sha256":"6ec34a9be2c1667437ff299162eef65e51c6e0b2b019bababe3525d7f4b44d8b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T02:04:53.645067Z","signature_b64":"1MVgxG7d4rwhhvjn9zunm0VYkkfLlodE5y4Bca9lidvkU9qAff7U3H72OPruWopvDhLf3kJl1YC7hTZ+e8SrDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1fcaa4e12a4d726ae141597aa34fa6ffc2e88baa8410f0e5a5dc797674fe4ae6","last_reissued_at":"2026-05-21T02:04:53.644059Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T02:04:53.644059Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Free-Grained Hierarchical Visual Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Hierarchical image recognition can learn consistent predictions from labels at any taxonomy level using mixed-granularity supervision.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Seulki Park, Stella X. Yu, Zilin Wang","submitted_at":"2025-10-16T14:35:18Z","abstract_excerpt":"Hierarchical image recognition seeks to predict class labels along a semantic taxonomy, from broad categories to specific ones, typically under the tidy assumption that every training image is fully annotated along its taxonomy path. Reality is messier: A distant bird may be labeled only bird, while a clear close-up may justify bald eagle. We introduce free-grain training, where labels may appear at any level of the taxonomy and models must learn consistent hierarchical predictions from incomplete, mixed-granularity supervision. We build benchmark datasets with varying label granularity and sh"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We introduce free-grain training, where labels may appear at any level of the taxonomy and models must learn consistent hierarchical predictions from incomplete, mixed-granularity supervision.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That supplementing with broad text-based supervision or framing missing labels as semi-supervised learning will sufficiently compensate for incomplete annotations without introducing new inconsistencies or performance drops.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Introduces free-grain hierarchical visual recognition with new benchmarks for mixed-granularity labels and methods using text-based supervision plus semi-supervised learning for missing labels.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Hierarchical image recognition can learn consistent predictions from labels at any taxonomy level using mixed-granularity supervision.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ecc2df2a5f39c7120a0138065adc5771733e0b994f9e75424c5c98de4a7eab79"},"source":{"id":"2510.14737","kind":"arxiv","version":3},"verdict":{"id":"9ccae7e9-e5b9-4c86-837d-1e96404f9df3","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T06:05:51.368233Z","strongest_claim":"We introduce free-grain training, where labels may appear at any level of the taxonomy and models must learn consistent hierarchical predictions from incomplete, mixed-granularity supervision.","one_line_summary":"Introduces free-grain hierarchical visual recognition with new benchmarks for mixed-granularity labels and methods using text-based supervision plus semi-supervised learning for missing labels.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That supplementing with broad text-based supervision or framing missing labels as semi-supervised learning will sufficiently compensate for incomplete annotations without introducing new inconsistencies or performance drops.","pith_extraction_headline":"Hierarchical image recognition can learn consistent predictions from labels at any taxonomy level using mixed-granularity supervision."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2510.14737/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":2,"snapshot_sha256":"bdb70b7f470661db12862d79bbed1a1bff4b3215c171514a6023c4b44de56179"},"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":"2510.14737","created_at":"2026-05-21T02:04:53.644201+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.14737v3","created_at":"2026-05-21T02:04:53.644201+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.14737","created_at":"2026-05-21T02:04:53.644201+00:00"},{"alias_kind":"pith_short_12","alias_value":"D7FKJYJKJVZG","created_at":"2026-05-21T02:04:53.644201+00:00"},{"alias_kind":"pith_short_16","alias_value":"D7FKJYJKJVZGVYKB","created_at":"2026-05-21T02:04:53.644201+00:00"},{"alias_kind":"pith_short_8","alias_value":"D7FKJYJK","created_at":"2026-05-21T02:04:53.644201+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77","json":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77.json","graph_json":"https://pith.science/api/pith-number/D7FKJYJKJVZGVYKBLF5KGT5G77/graph.json","events_json":"https://pith.science/api/pith-number/D7FKJYJKJVZGVYKBLF5KGT5G77/events.json","paper":"https://pith.science/paper/D7FKJYJK"},"agent_actions":{"view_html":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77","download_json":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77.json","view_paper":"https://pith.science/paper/D7FKJYJK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.14737&json=true","fetch_graph":"https://pith.science/api/pith-number/D7FKJYJKJVZGVYKBLF5KGT5G77/graph.json","fetch_events":"https://pith.science/api/pith-number/D7FKJYJKJVZGVYKBLF5KGT5G77/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77/action/storage_attestation","attest_author":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77/action/author_attestation","sign_citation":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77/action/citation_signature","submit_replication":"https://pith.science/pith/D7FKJYJKJVZGVYKBLF5KGT5G77/action/replication_record"}},"created_at":"2026-05-21T02:04:53.644201+00:00","updated_at":"2026-05-21T02:04:53.644201+00:00"}