{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OPOWPHNAHUAFW4EBE3HLIKGLUJ","short_pith_number":"pith:OPOWPHNA","schema_version":"1.0","canonical_sha256":"73dd679da03d005b708126ceb428cba241f9591dcb0ef192f72ed6e3901001ce","source":{"kind":"arxiv","id":"2605.13322","version":1},"attestation_state":"computed","paper":{"title":"KamonBench: A Grammar-Based Dataset for Evaluating Compositional Factor Recovery in Vision-Language Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"KamonBench supplies 20,000 grammar-generated crest images whose explicit container, modifier, and motif factors let models be scored directly on compositional recovery rather than caption match alone.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Richard Sproat, Stefano Peluchetti","submitted_at":"2026-05-13T10:35:07Z","abstract_excerpt":"Kamon (family crests) are an important part of Japanese culture and a natural test case for compositional visual recognition: each crest combines a small number of symbolic choices, but the space of possible descriptions is sparse. We introduce KamonBench, a grammar-based image-to-structure benchmark with 20,000 synthetic composite crests and auxiliary component examples. Each composite crest is paired with a formal kamon description language - \"kamon y\\=ogo\" - description, a segmented Japanese analysis, an English translation, and a non-linguistic program code. Because each synthetic crest is"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.13322","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T10:35:07Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"018246d88a61fe17bcde891e1739c69112d2d5f4e748bdc6510af481eee352ac","abstract_canon_sha256":"13de2adc754bafe5a892ae7bf65d2f8ae20a7bcc00474eb6a082691b611a5df6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:48.653115Z","signature_b64":"ySSGmbGPX7J20KUlMqNHmgZ8oHq1rxGphzZZ57C5tt/FN0RRKIuoqdBwGTMafO6m/troc7vkbsGcp+LhziWGAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"73dd679da03d005b708126ceb428cba241f9591dcb0ef192f72ed6e3901001ce","last_reissued_at":"2026-05-18T02:44:48.652722Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:48.652722Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"KamonBench: A Grammar-Based Dataset for Evaluating Compositional Factor Recovery in Vision-Language Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"KamonBench supplies 20,000 grammar-generated crest images whose explicit container, modifier, and motif factors let models be scored directly on compositional recovery rather than caption match alone.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Richard Sproat, Stefano Peluchetti","submitted_at":"2026-05-13T10:35:07Z","abstract_excerpt":"Kamon (family crests) are an important part of Japanese culture and a natural test case for compositional visual recognition: each crest combines a small number of symbolic choices, but the space of possible descriptions is sparse. We introduce KamonBench, a grammar-based image-to-structure benchmark with 20,000 synthetic composite crests and auxiliary component examples. Each composite crest is paired with a formal kamon description language - \"kamon y\\=ogo\" - description, a segmented Japanese analysis, an English translation, and a non-linguistic program code. Because each synthetic crest is"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"KamonBench therefore provides a controlled testbed for sparse compositional visual recognition and factor recovery in vision-language models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The grammar rules and synthetic generation process produce images whose factor structure mirrors the compositional challenges present in natural images and real-world visual recognition tasks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"KamonBench is a grammar-generated synthetic dataset of compositional kamon crests with explicit factor annotations to evaluate factor recovery in vision-language models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"KamonBench supplies 20,000 grammar-generated crest images whose explicit container, modifier, and motif factors let models be scored directly on compositional recovery rather than caption match alone.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"03e08ce8630d214ae1615d281ae62f130fb6aad4261d7be73377c19a18ab8634"},"source":{"id":"2605.13322","kind":"arxiv","version":1},"verdict":{"id":"d53b069c-3ca8-4f6a-a106-0a32150f6228","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:35:19.971470Z","strongest_claim":"KamonBench therefore provides a controlled testbed for sparse compositional visual recognition and factor recovery in vision-language models.","one_line_summary":"KamonBench is a grammar-generated synthetic dataset of compositional kamon crests with explicit factor annotations to evaluate factor recovery in vision-language models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The grammar rules and synthetic generation process produce images whose factor structure mirrors the compositional challenges present in natural images and real-world visual recognition tasks.","pith_extraction_headline":"KamonBench supplies 20,000 grammar-generated crest images whose explicit container, modifier, and motif factors let models be scored directly on compositional recovery rather than caption match alone."},"references":{"count":29,"sample":[{"doi":"","year":2017,"title":"Understanding intermediate layers using linear classi- fier probes","work_id":"9d32e99c-c7a8-4a12-8635-d27bff713b31","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1162/coli_a_00422","year":2022,"title":"Probing Classifiers: Promises, Shortcomings, and Advances","work_id":"3eabab74-ac71-4292-86ce-b0469cd4e6cf","ref_index":2,"cited_arxiv_id":"2102.12452","is_internal_anchor":true},{"doi":"","year":1993,"title":"Kadokawa Shoten (角川書店), Tokyo, 1993","work_id":"a978d7c1-72ab-4885-ad0b-94260b0a0831","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale","work_id":"e96730e3-129b-4db6-b981-15ab7932e297","ref_index":4,"cited_arxiv_id":"2010.11929","is_internal_anchor":true},{"doi":"","year":1971,"title":"John Weatherhill, New York, 1971","work_id":"ab8e0fd7-87a3-4a64-a7b3-76fa0a40bf07","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":29,"snapshot_sha256":"4b5904df336149eef7a9411004c35d6460abf6ef773bd5a1f539f40ce6248788","internal_anchors":4},"formal_canon":{"evidence_count":2,"snapshot_sha256":"aa62602e676d9ce1ef388edef9176981f836dd8864815dff05c0db49bdaed63b"},"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":"2605.13322","created_at":"2026-05-18T02:44:48.652792+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.13322v1","created_at":"2026-05-18T02:44:48.652792+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13322","created_at":"2026-05-18T02:44:48.652792+00:00"},{"alias_kind":"pith_short_12","alias_value":"OPOWPHNAHUAF","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"OPOWPHNAHUAFW4EB","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"OPOWPHNA","created_at":"2026-05-18T12:33:37.589309+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/OPOWPHNAHUAFW4EBE3HLIKGLUJ","json":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ.json","graph_json":"https://pith.science/api/pith-number/OPOWPHNAHUAFW4EBE3HLIKGLUJ/graph.json","events_json":"https://pith.science/api/pith-number/OPOWPHNAHUAFW4EBE3HLIKGLUJ/events.json","paper":"https://pith.science/paper/OPOWPHNA"},"agent_actions":{"view_html":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ","download_json":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ.json","view_paper":"https://pith.science/paper/OPOWPHNA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.13322&json=true","fetch_graph":"https://pith.science/api/pith-number/OPOWPHNAHUAFW4EBE3HLIKGLUJ/graph.json","fetch_events":"https://pith.science/api/pith-number/OPOWPHNAHUAFW4EBE3HLIKGLUJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ/action/storage_attestation","attest_author":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ/action/author_attestation","sign_citation":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ/action/citation_signature","submit_replication":"https://pith.science/pith/OPOWPHNAHUAFW4EBE3HLIKGLUJ/action/replication_record"}},"created_at":"2026-05-18T02:44:48.652792+00:00","updated_at":"2026-05-18T02:44:48.652792+00:00"}