{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:X2XLG4ZW24OSYHMLQ362KZWJUZ","short_pith_number":"pith:X2XLG4ZW","canonical_record":{"source":{"id":"2503.13652","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T18:59:29Z","cross_cats_sorted":[],"title_canon_sha256":"9656bbc6ab8f15075b326daef4e9091395e43b71ac0851ef0e4d004d60e9fd4b","abstract_canon_sha256":"f818f497dca0e24ada36e4dd9313b98eb9296872f5ec85a4535e6eeb8f96782a"},"schema_version":"1.0"},"canonical_sha256":"beaeb37336d71d2c1d8b86fda566c9a6597ded25d8a1328d6865f83844db34df","source":{"kind":"arxiv","id":"2503.13652","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.13652","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"arxiv_version","alias_value":"2503.13652v2","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.13652","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"pith_short_12","alias_value":"X2XLG4ZW24OS","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"pith_short_16","alias_value":"X2XLG4ZW24OSYHML","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"pith_short_8","alias_value":"X2XLG4ZW","created_at":"2026-07-05T11:47:20Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:X2XLG4ZW24OSYHMLQ362KZWJUZ","target":"record","payload":{"canonical_record":{"source":{"id":"2503.13652","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T18:59:29Z","cross_cats_sorted":[],"title_canon_sha256":"9656bbc6ab8f15075b326daef4e9091395e43b71ac0851ef0e4d004d60e9fd4b","abstract_canon_sha256":"f818f497dca0e24ada36e4dd9313b98eb9296872f5ec85a4535e6eeb8f96782a"},"schema_version":"1.0"},"canonical_sha256":"beaeb37336d71d2c1d8b86fda566c9a6597ded25d8a1328d6865f83844db34df","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:47:20.538371Z","signature_b64":"0yS+bV0sCrT6jTLxypybB4ZTzWaRWqflPHrNK1fWwMEz3TI87+K54P88VjsO4Abi4qEkeHptifMDqwQxOZjRBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"beaeb37336d71d2c1d8b86fda566c9a6597ded25d8a1328d6865f83844db34df","last_reissued_at":"2026-07-05T11:47:20.537877Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:47:20.537877Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2503.13652","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:47:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"43t2BWVbTXZJduqITAj7xX2n7gEivX8w0dEniRyALEIf/W392v+bQb1dQvRlEbxX2uclIOksSKKppdNAmaxjAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-06T18:46:31.354330Z"},"content_sha256":"311828c892543203ecff606360656efc53d990005ac9298251d7a275e224be1a","schema_version":"1.0","event_id":"sha256:311828c892543203ecff606360656efc53d990005ac9298251d7a275e224be1a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:X2XLG4ZW24OSYHMLQ362KZWJUZ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Web Artifact Attacks Disrupt Vision Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bryan A. Plummer, Kate Saenko, Maan Qraitem, Piotr Teterwak","submitted_at":"2025-03-17T18:59:29Z","abstract_excerpt":"Vision-language models (VLMs) (e.g. CLIP, LLaVA) are trained on large-scale, lightly curated web datasets, leading them to learn unintended correlations between semantic concepts and unrelated visual signals. These associations degrade model accuracy by causing predictions to rely on incidental patterns rather than genuine visual understanding. Prior work has weaponized these correlations as an attack vector to manipulate model predictions, such as inserting a deceiving class text onto the image in a \"typographic\" attack. These attacks succeed due to VLMs' text-heavy bias-a result of captions "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.13652","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/2503.13652/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:47:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LwQNnqCv4HdgYp/Tms8mSc8ESi//gDdzOHqWGmATeXCoemi8k1sXnwqVZ+fnj8tPkX1abe9O3jVRyhkGl+nSBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-06T18:46:31.354710Z"},"content_sha256":"d6e55cf907174fa679ca43860955f4a8578ddc434874f6574b287853f7ce6f94","schema_version":"1.0","event_id":"sha256:d6e55cf907174fa679ca43860955f4a8578ddc434874f6574b287853f7ce6f94"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ/bundle.json","state_url":"https://pith.science/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-06T18:46:31Z","links":{"resolver":"https://pith.science/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ","bundle":"https://pith.science/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ/bundle.json","state":"https://pith.science/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/X2XLG4ZW24OSYHMLQ362KZWJUZ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:X2XLG4ZW24OSYHMLQ362KZWJUZ","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f818f497dca0e24ada36e4dd9313b98eb9296872f5ec85a4535e6eeb8f96782a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T18:59:29Z","title_canon_sha256":"9656bbc6ab8f15075b326daef4e9091395e43b71ac0851ef0e4d004d60e9fd4b"},"schema_version":"1.0","source":{"id":"2503.13652","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.13652","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"arxiv_version","alias_value":"2503.13652v2","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.13652","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"pith_short_12","alias_value":"X2XLG4ZW24OS","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"pith_short_16","alias_value":"X2XLG4ZW24OSYHML","created_at":"2026-07-05T11:47:20Z"},{"alias_kind":"pith_short_8","alias_value":"X2XLG4ZW","created_at":"2026-07-05T11:47:20Z"}],"graph_snapshots":[{"event_id":"sha256:d6e55cf907174fa679ca43860955f4a8578ddc434874f6574b287853f7ce6f94","target":"graph","created_at":"2026-07-05T11:47:20Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2503.13652/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Vision-language models (VLMs) (e.g. CLIP, LLaVA) are trained on large-scale, lightly curated web datasets, leading them to learn unintended correlations between semantic concepts and unrelated visual signals. These associations degrade model accuracy by causing predictions to rely on incidental patterns rather than genuine visual understanding. Prior work has weaponized these correlations as an attack vector to manipulate model predictions, such as inserting a deceiving class text onto the image in a \"typographic\" attack. These attacks succeed due to VLMs' text-heavy bias-a result of captions ","authors_text":"Bryan A. Plummer, Kate Saenko, Maan Qraitem, Piotr Teterwak","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T18:59:29Z","title":"Web Artifact Attacks Disrupt Vision Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.13652","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:311828c892543203ecff606360656efc53d990005ac9298251d7a275e224be1a","target":"record","created_at":"2026-07-05T11:47:20Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f818f497dca0e24ada36e4dd9313b98eb9296872f5ec85a4535e6eeb8f96782a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-17T18:59:29Z","title_canon_sha256":"9656bbc6ab8f15075b326daef4e9091395e43b71ac0851ef0e4d004d60e9fd4b"},"schema_version":"1.0","source":{"id":"2503.13652","kind":"arxiv","version":2}},"canonical_sha256":"beaeb37336d71d2c1d8b86fda566c9a6597ded25d8a1328d6865f83844db34df","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"beaeb37336d71d2c1d8b86fda566c9a6597ded25d8a1328d6865f83844db34df","first_computed_at":"2026-07-05T11:47:20.537877Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:47:20.537877Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"0yS+bV0sCrT6jTLxypybB4ZTzWaRWqflPHrNK1fWwMEz3TI87+K54P88VjsO4Abi4qEkeHptifMDqwQxOZjRBw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:47:20.538371Z","signed_message":"canonical_sha256_bytes"},"source_id":"2503.13652","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:311828c892543203ecff606360656efc53d990005ac9298251d7a275e224be1a","sha256:d6e55cf907174fa679ca43860955f4a8578ddc434874f6574b287853f7ce6f94"],"state_sha256":"41fd6123dc544044e46bca47772993832fcbbd1fb94bdec990a59a2e5b31b3af"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AbrHS/DqtDw7XmcPRMzMil+OV4bFzVd/H7O1GGNQOGSR4T/ZHo0w4JPq6IrzupqmDNncnEbwYrVT2cF4mCdCCQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-06T18:46:31.356660Z","bundle_sha256":"9c3449f38520278b8980f29a3cc950d12c8e4144a7e95f73ab415f7c5c6a7a55"}}