{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2RNBAQV3RP7EDPOZFTYRWLXUOW","short_pith_number":"pith:2RNBAQV3","schema_version":"1.0","canonical_sha256":"d45a1042bb8bfe41bdd92cf11b2ef475a66d5a775952f43b0eb8b6528bcf1a61","source":{"kind":"arxiv","id":"2607.13305","version":1},"attestation_state":"computed","paper":{"title":"Accuracy Without Grounding: Diagnosing Visual Dependency Dissociation in Video LLM Benchmarks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.MM"],"primary_cat":"cs.CV","authors_text":"Jae Joong Lee","submitted_at":"2026-07-14T22:15:50Z","abstract_excerpt":"Benchmark accuracy in video large language models (LLMs) is often treated as evidence of visual understanding. We audit this assumption across twenty models spanning 2-78B parameters and ten architecture families. We introduce the Visual Dependency Gap (VDG), the difference in per-question correctness between original-video and black-screen conditions. Paired McNemar tests on MVBench show that accuracy and visual dependency are separable: models differ on original video (p = 0.0003) but not on black screens (p = 0.53). Across models, task-type rankings are stable: Attribute Perception is stron"},"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":"2607.13305","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-14T22:15:50Z","cross_cats_sorted":["cs.AI","cs.MM"],"title_canon_sha256":"32b828dcf1b27ee00acb9536db292b0983b61cada91234f67275d861d34f918d","abstract_canon_sha256":"06b0693c4422cfd76939a9e2fd0ab11c1838e1b0b8cc2a4a9d75e9cb665ec3fe"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-16T00:22:11.277799Z","signature_b64":"UyFIOmiFsapo/fMOz+rwHfCaSVX9l0oBWniiETaZlzIc86QU3LdQ9h15ibRX1PwD9vMWSCoBJlSneugmZSwUDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d45a1042bb8bfe41bdd92cf11b2ef475a66d5a775952f43b0eb8b6528bcf1a61","last_reissued_at":"2026-07-16T00:22:11.276956Z","signature_status":"signed_v1","first_computed_at":"2026-07-16T00:22:11.276956Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accuracy Without Grounding: Diagnosing Visual Dependency Dissociation in Video LLM Benchmarks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.MM"],"primary_cat":"cs.CV","authors_text":"Jae Joong Lee","submitted_at":"2026-07-14T22:15:50Z","abstract_excerpt":"Benchmark accuracy in video large language models (LLMs) is often treated as evidence of visual understanding. We audit this assumption across twenty models spanning 2-78B parameters and ten architecture families. We introduce the Visual Dependency Gap (VDG), the difference in per-question correctness between original-video and black-screen conditions. Paired McNemar tests on MVBench show that accuracy and visual dependency are separable: models differ on original video (p = 0.0003) but not on black screens (p = 0.53). Across models, task-type rankings are stable: Attribute Perception is stron"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.13305","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/2607.13305/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":"2607.13305","created_at":"2026-07-16T00:22:11.277399+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.13305v1","created_at":"2026-07-16T00:22:11.277399+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.13305","created_at":"2026-07-16T00:22:11.277399+00:00"},{"alias_kind":"pith_short_12","alias_value":"2RNBAQV3RP7E","created_at":"2026-07-16T00:22:11.277399+00:00"},{"alias_kind":"pith_short_16","alias_value":"2RNBAQV3RP7EDPOZ","created_at":"2026-07-16T00:22:11.277399+00:00"},{"alias_kind":"pith_short_8","alias_value":"2RNBAQV3","created_at":"2026-07-16T00:22:11.277399+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/2RNBAQV3RP7EDPOZFTYRWLXUOW","json":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW.json","graph_json":"https://pith.science/api/pith-number/2RNBAQV3RP7EDPOZFTYRWLXUOW/graph.json","events_json":"https://pith.science/api/pith-number/2RNBAQV3RP7EDPOZFTYRWLXUOW/events.json","paper":"https://pith.science/paper/2RNBAQV3"},"agent_actions":{"view_html":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW","download_json":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW.json","view_paper":"https://pith.science/paper/2RNBAQV3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.13305&json=true","fetch_graph":"https://pith.science/api/pith-number/2RNBAQV3RP7EDPOZFTYRWLXUOW/graph.json","fetch_events":"https://pith.science/api/pith-number/2RNBAQV3RP7EDPOZFTYRWLXUOW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW/action/storage_attestation","attest_author":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW/action/author_attestation","sign_citation":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW/action/citation_signature","submit_replication":"https://pith.science/pith/2RNBAQV3RP7EDPOZFTYRWLXUOW/action/replication_record"}},"created_at":"2026-07-16T00:22:11.277399+00:00","updated_at":"2026-07-16T00:22:11.277399+00:00"}