{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XMXDSXEDPTLLBOKOQT6QPPVH25","short_pith_number":"pith:XMXDSXED","schema_version":"1.0","canonical_sha256":"bb2e395c837cd6b0b94e84fd07bea7d74e2d89674bc99a7130a7634b0eceafd6","source":{"kind":"arxiv","id":"2605.15325","version":1},"attestation_state":"computed","paper":{"title":"COPRA: Conditional Parameter Adaptation with Reinforcement Learning for Video Anomaly Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"COPRA uses reinforcement learning to generate input-specific parameter updates that dynamically adapt a frozen vision-language model to each video segment for anomaly detection.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Darryl Cherian Jacob, Kai Wang, Pan He, Xinyu Liu","submitted_at":"2026-05-14T18:39:40Z","abstract_excerpt":"Vision-language models (VLMs) have shown strong performance in video anomaly detection (VAD) while providing interpretable predictions. However, existing VLM-based VAD methods suffer from a fundamental mismatch between training and inference in both data distribution and model configuration. First, most approaches rely on static post-training adaptation, limiting generalization under distribution shifts such as unseen environments or anomaly types. Second, they train VLMs on sparse frames from long videos, but perform inference on densely sampled short segments, creating inconsistencies betwee"},"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":"2605.15325","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T18:39:40Z","cross_cats_sorted":[],"title_canon_sha256":"9fb6019da126079b02b8937f97646fd00ef4007e0966a848ebdb95701d8802ed","abstract_canon_sha256":"0193d3fc29c3f044cdd902aae0c909281da8816d4d33468cee0b29c715566720"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:00:52.733629Z","signature_b64":"LfXio5HKPodBP0DMOJ3I5W7/7Yoz78Y9mIe0B8yE5P0Q0zPzUEfb6SbMjS4riGuYjV882K7f0JZCGydiBbfvCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bb2e395c837cd6b0b94e84fd07bea7d74e2d89674bc99a7130a7634b0eceafd6","last_reissued_at":"2026-05-20T00:00:52.732906Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:00:52.732906Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"COPRA: Conditional Parameter Adaptation with Reinforcement Learning for Video Anomaly Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"COPRA uses reinforcement learning to generate input-specific parameter updates that dynamically adapt a frozen vision-language model to each video segment for anomaly detection.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Darryl Cherian Jacob, Kai Wang, Pan He, Xinyu Liu","submitted_at":"2026-05-14T18:39:40Z","abstract_excerpt":"Vision-language models (VLMs) have shown strong performance in video anomaly detection (VAD) while providing interpretable predictions. However, existing VLM-based VAD methods suffer from a fundamental mismatch between training and inference in both data distribution and model configuration. First, most approaches rely on static post-training adaptation, limiting generalization under distribution shifts such as unseen environments or anomaly types. Second, they train VLMs on sparse frames from long videos, but perform inference on densely sampled short segments, creating inconsistencies betwee"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"COPRA generates input-specific parameter updates to dynamically adapt a frozen VLM for each video segment during both training and inference, consistently outperforming static baselines in both in-domain and cross-domain settings and generalizing to unseen tasks such as multiple-choice Video Question Answering and Dense Captioning.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That reinforcement learning can stably and effectively produce useful input-conditioned parameter updates for a frozen VLM without requiring domain-specific hyperparameter search or suffering from instability when applied to new video distributions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"COPRA introduces conditional parameter adaptation via RL to dynamically tune frozen VLMs for video anomaly detection, outperforming static methods in in-domain and cross-domain settings while generalizing to other video tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"COPRA uses reinforcement learning to generate input-specific parameter updates that dynamically adapt a frozen vision-language model to each video segment for anomaly detection.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"59bcb068731411502ab9dc7c7d5ef5e53291bed1a2f1d3e4300fa607f96e922b"},"source":{"id":"2605.15325","kind":"arxiv","version":1},"verdict":{"id":"6861777c-4515-4dcb-9575-96fdad5a3dfd","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T16:09:32.571812Z","strongest_claim":"COPRA generates input-specific parameter updates to dynamically adapt a frozen VLM for each video segment during both training and inference, consistently outperforming static baselines in both in-domain and cross-domain settings and generalizing to unseen tasks such as multiple-choice Video Question Answering and Dense Captioning.","one_line_summary":"COPRA introduces conditional parameter adaptation via RL to dynamically tune frozen VLMs for video anomaly detection, outperforming static methods in in-domain and cross-domain settings while generalizing to other video tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That reinforcement learning can stably and effectively produce useful input-conditioned parameter updates for a frozen VLM without requiring domain-specific hyperparameter search or suffering from instability when applied to new video distributions.","pith_extraction_headline":"COPRA uses reinforcement learning to generate input-specific parameter updates that dynamically adapt a frozen vision-language model to each video segment for anomaly detection."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15325/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:31:18.297134Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T16:16:04.938979Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.201027Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.765728Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"ffe681358e6c44af1ca09be95830ca7a124d289c784ac065037f2f3e71c206ee"},"references":{"count":64,"sample":[{"doi":"","year":null,"title":"Unlocking vision-language models for video anomaly detection via fine-grained prompting , author=. WACV , pages=","work_id":"b0aae5c3-438e-4059-9edd-2aa1c112a2bf","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Real-world anomaly detection in surveillance videos , author=. CVPR , pages=","work_id":"972a4ceb-15ea-40f4-8f05-aea86025fd26","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Workshop on Neural Network Weights as a New Data Modality , year=","work_id":"07eceb93-1ebd-4b76-8763-16b54874d7d4","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks , author=. CVPR , pages=","work_id":"4235ae3c-9daf-43a6-a196-d5acefb3ac05","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"A Survey of Weight Space Learning: Understanding, Representation, and Generation , author=. 2026 , eprint=","work_id":"ba86c074-5032-4554-bd61-e8226a957e26","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":64,"snapshot_sha256":"eb0e9971d1f9cfa53664cee409881ec15031717fcad1283661c0c22660da1c34","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1f0e61c55280db211b1548cfdd458f60376f3030a15f192731b57fea6a9cb6d0"},"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":"2605.15325","created_at":"2026-05-20T00:00:52.733017+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15325v1","created_at":"2026-05-20T00:00:52.733017+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15325","created_at":"2026-05-20T00:00:52.733017+00:00"},{"alias_kind":"pith_short_12","alias_value":"XMXDSXEDPTLL","created_at":"2026-05-20T00:00:52.733017+00:00"},{"alias_kind":"pith_short_16","alias_value":"XMXDSXEDPTLLBOKO","created_at":"2026-05-20T00:00:52.733017+00:00"},{"alias_kind":"pith_short_8","alias_value":"XMXDSXED","created_at":"2026-05-20T00:00:52.733017+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/XMXDSXEDPTLLBOKOQT6QPPVH25","json":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25.json","graph_json":"https://pith.science/api/pith-number/XMXDSXEDPTLLBOKOQT6QPPVH25/graph.json","events_json":"https://pith.science/api/pith-number/XMXDSXEDPTLLBOKOQT6QPPVH25/events.json","paper":"https://pith.science/paper/XMXDSXED"},"agent_actions":{"view_html":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25","download_json":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25.json","view_paper":"https://pith.science/paper/XMXDSXED","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15325&json=true","fetch_graph":"https://pith.science/api/pith-number/XMXDSXEDPTLLBOKOQT6QPPVH25/graph.json","fetch_events":"https://pith.science/api/pith-number/XMXDSXEDPTLLBOKOQT6QPPVH25/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25/action/storage_attestation","attest_author":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25/action/author_attestation","sign_citation":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25/action/citation_signature","submit_replication":"https://pith.science/pith/XMXDSXEDPTLLBOKOQT6QPPVH25/action/replication_record"}},"created_at":"2026-05-20T00:00:52.733017+00:00","updated_at":"2026-05-20T00:00:52.733017+00:00"}