{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HVASJFYU7UEO7HUDUXCKCFY2GE","short_pith_number":"pith:HVASJFYU","schema_version":"1.0","canonical_sha256":"3d41249714fd08ef9e83a5c4a1171a31121cc138c0b56e05eb94aa5e63beb022","source":{"kind":"arxiv","id":"2607.14256","version":1},"attestation_state":"computed","paper":{"title":"Automatic Hard Example Synthesis with Multi-Level Agentic Data Curation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.MA"],"primary_cat":"cs.AI","authors_text":"Ariel Fuxman, Bla\\v{z} Bratani\\v{c}, Genglin Liu, Krishnamurthy Viswanathan, Muye Zhang, Nathan L Clement, Nichole J. Hansen, Shalini Ghosh","submitted_at":"2026-07-15T18:13:05Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) are increasingly deployed for nuanced content safety and moderation tasks, yet they remain vulnerable to adversarial attacks and out-of-distribution edge cases. Traditional active learning and manual annotation fail to scale against the complexity and volume of novel multimodal threats. In this paper, we propose an automated, agentic red-teaming framework that systematically synthesizes difficult examples using an iterative strategy that proposes novel hypotheses as well as mutating on past attempts. Leveraging a multi-agent architecture that consists o"},"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":"2607.14256","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-15T18:13:05Z","cross_cats_sorted":["cs.MA"],"title_canon_sha256":"3b4b37e679309ead2e82d91c9dd5eba260a7fa2ede91d4dc586145222258016d","abstract_canon_sha256":"152bd1045af7a5116a4ee4ce85de4a9177236341959e62db215f4fe5039ee142"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T00:21:01.831037Z","signature_b64":"azs00P5MXPusuBt0iBQR2dCRlLk9d8+9LFyad4vrWzXUjbR3KtlHwkVAMiAxcoCJWlomilS53jLwus+bFX8OBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d41249714fd08ef9e83a5c4a1171a31121cc138c0b56e05eb94aa5e63beb022","last_reissued_at":"2026-07-17T00:21:01.830180Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T00:21:01.830180Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Automatic Hard Example Synthesis with Multi-Level Agentic Data Curation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.MA"],"primary_cat":"cs.AI","authors_text":"Ariel Fuxman, Bla\\v{z} Bratani\\v{c}, Genglin Liu, Krishnamurthy Viswanathan, Muye Zhang, Nathan L Clement, Nichole J. Hansen, Shalini Ghosh","submitted_at":"2026-07-15T18:13:05Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) are increasingly deployed for nuanced content safety and moderation tasks, yet they remain vulnerable to adversarial attacks and out-of-distribution edge cases. Traditional active learning and manual annotation fail to scale against the complexity and volume of novel multimodal threats. In this paper, we propose an automated, agentic red-teaming framework that systematically synthesizes difficult examples using an iterative strategy that proposes novel hypotheses as well as mutating on past attempts. Leveraging a multi-agent architecture that consists o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14256","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/2607.14256/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":"2607.14256","created_at":"2026-07-17T00:21:01.830633+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14256v1","created_at":"2026-07-17T00:21:01.830633+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14256","created_at":"2026-07-17T00:21:01.830633+00:00"},{"alias_kind":"pith_short_12","alias_value":"HVASJFYU7UEO","created_at":"2026-07-17T00:21:01.830633+00:00"},{"alias_kind":"pith_short_16","alias_value":"HVASJFYU7UEO7HUD","created_at":"2026-07-17T00:21:01.830633+00:00"},{"alias_kind":"pith_short_8","alias_value":"HVASJFYU","created_at":"2026-07-17T00:21:01.830633+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/HVASJFYU7UEO7HUDUXCKCFY2GE","json":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE.json","graph_json":"https://pith.science/api/pith-number/HVASJFYU7UEO7HUDUXCKCFY2GE/graph.json","events_json":"https://pith.science/api/pith-number/HVASJFYU7UEO7HUDUXCKCFY2GE/events.json","paper":"https://pith.science/paper/HVASJFYU"},"agent_actions":{"view_html":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE","download_json":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE.json","view_paper":"https://pith.science/paper/HVASJFYU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14256&json=true","fetch_graph":"https://pith.science/api/pith-number/HVASJFYU7UEO7HUDUXCKCFY2GE/graph.json","fetch_events":"https://pith.science/api/pith-number/HVASJFYU7UEO7HUDUXCKCFY2GE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE/action/storage_attestation","attest_author":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE/action/author_attestation","sign_citation":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE/action/citation_signature","submit_replication":"https://pith.science/pith/HVASJFYU7UEO7HUDUXCKCFY2GE/action/replication_record"}},"created_at":"2026-07-17T00:21:01.830633+00:00","updated_at":"2026-07-17T00:21:01.830633+00:00"}