{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:N4TABZ4CK6HTPNMWYVUZUCXTIG","short_pith_number":"pith:N4TABZ4C","schema_version":"1.0","canonical_sha256":"6f2600e782578f37b596c5699a0af341b6c80b1fe99a003dafb89e4f22dee581","source":{"kind":"arxiv","id":"2412.06878","version":1},"attestation_state":"computed","paper":{"title":"SafeWatch: An Efficient Safety-Policy Following Video Guardrail Model with Transparent Explanations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Bo Li, Francesco Pinto, Minzhou Pan, Zhaorun Chen","submitted_at":"2024-12-09T18:59:04Z","abstract_excerpt":"With the rise of generative AI and rapid growth of high-quality video generation, video guardrails have become more crucial than ever to ensure safety and security across platforms. Current video guardrails, however, are either overly simplistic, relying on pure classification models trained on simple policies with limited unsafe categories, which lack detailed explanations, or prompting multimodal large language models (MLLMs) with long safety guidelines, which are inefficient and impractical for guardrailing real-world content. To bridge this gap, we propose SafeWatch, an efficient MLLM-base"},"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":"2412.06878","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-12-09T18:59:04Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"861ba75f337e62f810b95b73499e2f8b03057e1fd7712f348d9df15024428629","abstract_canon_sha256":"897956a302ca219cf534054915cb20213aeb8abc53dd02263c914279a9a82986"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:46:59.167935Z","signature_b64":"n6LBSezQ8vOamPW8GPckb68i+SrI2nVoAlhzZ/RZigSFV3la4gBCKPv0oPbR/2kZ2hXRkxuoncp1eoa9H8kfBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6f2600e782578f37b596c5699a0af341b6c80b1fe99a003dafb89e4f22dee581","last_reissued_at":"2026-07-05T09:46:59.167333Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:46:59.167333Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeWatch: An Efficient Safety-Policy Following Video Guardrail Model with Transparent Explanations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Bo Li, Francesco Pinto, Minzhou Pan, Zhaorun Chen","submitted_at":"2024-12-09T18:59:04Z","abstract_excerpt":"With the rise of generative AI and rapid growth of high-quality video generation, video guardrails have become more crucial than ever to ensure safety and security across platforms. Current video guardrails, however, are either overly simplistic, relying on pure classification models trained on simple policies with limited unsafe categories, which lack detailed explanations, or prompting multimodal large language models (MLLMs) with long safety guidelines, which are inefficient and impractical for guardrailing real-world content. To bridge this gap, we propose SafeWatch, an efficient MLLM-base"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.06878","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/2412.06878/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":"2412.06878","created_at":"2026-07-05T09:46:59.167397+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.06878v1","created_at":"2026-07-05T09:46:59.167397+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.06878","created_at":"2026-07-05T09:46:59.167397+00:00"},{"alias_kind":"pith_short_12","alias_value":"N4TABZ4CK6HT","created_at":"2026-07-05T09:46:59.167397+00:00"},{"alias_kind":"pith_short_16","alias_value":"N4TABZ4CK6HTPNMW","created_at":"2026-07-05T09:46:59.167397+00:00"},{"alias_kind":"pith_short_8","alias_value":"N4TABZ4C","created_at":"2026-07-05T09:46:59.167397+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.23883","citing_title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","ref_index":210,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01761","citing_title":"TrajShield: Trajectory-Level Safety Mediation for Defending Text-to-Video Models Against Jailbreak Attacks","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG","json":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG.json","graph_json":"https://pith.science/api/pith-number/N4TABZ4CK6HTPNMWYVUZUCXTIG/graph.json","events_json":"https://pith.science/api/pith-number/N4TABZ4CK6HTPNMWYVUZUCXTIG/events.json","paper":"https://pith.science/paper/N4TABZ4C"},"agent_actions":{"view_html":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG","download_json":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG.json","view_paper":"https://pith.science/paper/N4TABZ4C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.06878&json=true","fetch_graph":"https://pith.science/api/pith-number/N4TABZ4CK6HTPNMWYVUZUCXTIG/graph.json","fetch_events":"https://pith.science/api/pith-number/N4TABZ4CK6HTPNMWYVUZUCXTIG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG/action/storage_attestation","attest_author":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG/action/author_attestation","sign_citation":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG/action/citation_signature","submit_replication":"https://pith.science/pith/N4TABZ4CK6HTPNMWYVUZUCXTIG/action/replication_record"}},"created_at":"2026-07-05T09:46:59.167397+00:00","updated_at":"2026-07-05T09:46:59.167397+00:00"}