{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:BKGNKZ6WUAON6SUTGVSBZYDWBU","short_pith_number":"pith:BKGNKZ6W","canonical_record":{"source":{"id":"2605.08810","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:02:20Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"cbb19f3fe370e9bbabbc402f222d56c4e3b03826415b17c7d56c430daf44ba42","abstract_canon_sha256":"145c87945dfcaf56feb9ff6429aec15b39ce6a3d310d237851f217807a285039"},"schema_version":"1.0"},"canonical_sha256":"0a8cd567d6a01cdf4a9335641ce0760d0e5bd3457bb269e5c53921baf3db4222","source":{"kind":"arxiv","id":"2605.08810","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.08810","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"arxiv_version","alias_value":"2605.08810v2","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08810","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"pith_short_12","alias_value":"BKGNKZ6WUAON","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"pith_short_16","alias_value":"BKGNKZ6WUAON6SUT","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"pith_short_8","alias_value":"BKGNKZ6W","created_at":"2026-07-30T01:17:00Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:BKGNKZ6WUAON6SUTGVSBZYDWBU","target":"record","payload":{"canonical_record":{"source":{"id":"2605.08810","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:02:20Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"cbb19f3fe370e9bbabbc402f222d56c4e3b03826415b17c7d56c430daf44ba42","abstract_canon_sha256":"145c87945dfcaf56feb9ff6429aec15b39ce6a3d310d237851f217807a285039"},"schema_version":"1.0"},"canonical_sha256":"0a8cd567d6a01cdf4a9335641ce0760d0e5bd3457bb269e5c53921baf3db4222","receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0a8cd567d6a01cdf4a9335641ce0760d0e5bd3457bb269e5c53921baf3db4222","last_reissued_at":"2026-07-30T01:17:00.268214Z","signature_status":"unsigned_v0","first_computed_at":"2026-07-30T01:17:00.268214Z"},"source_kind":"arxiv","source_id":"2605.08810","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":"2026-07-30T01:17:00Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"11SEYxasFzNP8WtUD/HKB2nlgAjBSS3INAzMXQA7Wf9KKAVvmmLs1Xg2gu6KXvevBCadyWk/3yMh9TRL51+/Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-14T08:14:16.505022Z"},"content_sha256":"556fa5a873637bc0b731c8cd3d57ed3df3c4f59461937d8d0d9d9247612d7f8d","schema_version":"1.0","event_id":"sha256:556fa5a873637bc0b731c8cd3d57ed3df3c4f59461937d8d0d9d9247612d7f8d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:BKGNKZ6WUAON6SUTGVSBZYDWBU","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Compressed Video Aggregator: Content-driven Module for Efficient Micro-Video Recommendation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"The Compressed Video Aggregator decouples video content from preference learning by combining frozen VFM embeddings through latent reasoning to create compact representations that improve micro-video recommendation accuracy while cutting训练","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bo Hui, Chao Jiang, Fei Wang, Huiyuan Chen, Kaiyuan Deng, Ruimeng Ye, Xiaolong Ma, Yang Xiao, Zinan Ling","submitted_at":"2026-05-09T09:02:20Z","abstract_excerpt":"We propose \\textbf{Compressed Video Aggregator} (CVA), a lightweight micro-video recommendation module that decouples video information from preference learning. CVA first summarizes frozen VFM frame embeddings into a semantic-consensus anchor through masked mean pooling, projects this anchor into a compact latent space, and refines the projected representation with residual self-attention and feedforward blocks before producing a single video embedding for the recommender. Due to the redundancy in the frame count of the original benchmark and its overly coarse sampling, we used titles to re-s"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CVA aggregates frozen VFM embeddings, and uses latent reasoning without cross-attention projection, producing compact video embeddings for recommenders. Experiments show consistent gains with orders-of-magnitude reductions in training time and GPU memory; re-selected frames further enhance performance of all methods.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That embeddings from frozen VFMs plus title-based CLIP key-frame selection capture sufficient video content for preference learning without requiring end-to-end fine-tuning or full video processing.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CVA aggregates frozen VFM embeddings via latent reasoning to create compact video embeddings for efficient micro-video recommendation, delivering consistent performance gains and orders-of-magnitude efficiency improvements.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"The Compressed Video Aggregator decouples video content from preference learning by combining frozen VFM embeddings through latent reasoning to create compact representations that improve micro-video recommendation accuracy while cutting训练","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"52868cb76e50dee1d24d69cbd753aad465c09bac1da136e635c28ae863c413b4"},"source":{"id":"2605.08810","kind":"arxiv","version":2},"verdict":{"id":"20d756c2-38bb-43f8-b910-de65e2c04612","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T02:05:09.875128Z","strongest_claim":"CVA aggregates frozen VFM embeddings, and uses latent reasoning without cross-attention projection, producing compact video embeddings for recommenders. Experiments show consistent gains with orders-of-magnitude reductions in training time and GPU memory; re-selected frames further enhance performance of all methods.","one_line_summary":"CVA aggregates frozen VFM embeddings via latent reasoning to create compact video embeddings for efficient micro-video recommendation, delivering consistent performance gains and orders-of-magnitude efficiency improvements.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That embeddings from frozen VFMs plus title-based CLIP key-frame selection capture sufficient video content for preference learning without requiring end-to-end fine-tuning or full video processing.","pith_extraction_headline":"The Compressed Video Aggregator decouples video content from preference learning by combining frozen VFM embeddings through latent reasoning to create compact representations that improve micro-video recommendation accuracy while cutting训练"},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.08810/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T08:42:02.046603Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T22:34:19.606198Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T14:01:21.969607Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T10:48:49.676665Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"a8c3cbf8b4cf319ab3517a86948db6f7349aa7c1fb01b9da72ae289289591a9f"},"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":"20d756c2-38bb-43f8-b910-de65e2c04612"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-30T01:17:00Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ThK2zCUt5VYJcLwwpmxqy4WTYrVMWZd7/ZUSTlOOEBbKkedivoB9cXWwqrALeEAGbp4i3jqUBo9h1ALJ5LR/Bg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-14T08:14:16.506171Z"},"content_sha256":"c70466493e79ac3062f3a33b852fae5313138b0bb5cb82c699ce0df4bb641e61","schema_version":"1.0","event_id":"sha256:c70466493e79ac3062f3a33b852fae5313138b0bb5cb82c699ce0df4bb641e61"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU/bundle.json","state_url":"https://pith.science/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU/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":"2026-08-14T08:14:16Z","links":{"resolver":"https://pith.science/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU","bundle":"https://pith.science/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU/bundle.json","state":"https://pith.science/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU/state.json","well_known_bundle":"https://pith.science/.well-known/pith/BKGNKZ6WUAON6SUTGVSBZYDWBU/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:BKGNKZ6WUAON6SUTGVSBZYDWBU","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":"145c87945dfcaf56feb9ff6429aec15b39ce6a3d310d237851f217807a285039","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:02:20Z","title_canon_sha256":"cbb19f3fe370e9bbabbc402f222d56c4e3b03826415b17c7d56c430daf44ba42"},"schema_version":"1.0","source":{"id":"2605.08810","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.08810","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"arxiv_version","alias_value":"2605.08810v2","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08810","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"pith_short_12","alias_value":"BKGNKZ6WUAON","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"pith_short_16","alias_value":"BKGNKZ6WUAON6SUT","created_at":"2026-07-30T01:17:00Z"},{"alias_kind":"pith_short_8","alias_value":"BKGNKZ6W","created_at":"2026-07-30T01:17:00Z"}],"graph_snapshots":[{"event_id":"sha256:c70466493e79ac3062f3a33b852fae5313138b0bb5cb82c699ce0df4bb641e61","target":"graph","created_at":"2026-07-30T01:17:00Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"CVA aggregates frozen VFM embeddings, and uses latent reasoning without cross-attention projection, producing compact video embeddings for recommenders. Experiments show consistent gains with orders-of-magnitude reductions in training time and GPU memory; re-selected frames further enhance performance of all methods."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That embeddings from frozen VFMs plus title-based CLIP key-frame selection capture sufficient video content for preference learning without requiring end-to-end fine-tuning or full video processing."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"CVA aggregates frozen VFM embeddings via latent reasoning to create compact video embeddings for efficient micro-video recommendation, delivering consistent performance gains and orders-of-magnitude efficiency improvements."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"The Compressed Video Aggregator decouples video content from preference learning by combining frozen VFM embeddings through latent reasoning to create compact representations that improve micro-video recommendation accuracy while cutting训练"}],"snapshot_sha256":"52868cb76e50dee1d24d69cbd753aad465c09bac1da136e635c28ae863c413b4"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-20T08:42:02.046603Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T22:34:19.606198Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T14:01:21.969607Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T10:48:49.676665Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.08810/integrity.json","findings":[],"snapshot_sha256":"a8c3cbf8b4cf319ab3517a86948db6f7349aa7c1fb01b9da72ae289289591a9f","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We propose \\textbf{Compressed Video Aggregator} (CVA), a lightweight micro-video recommendation module that decouples video information from preference learning. CVA first summarizes frozen VFM frame embeddings into a semantic-consensus anchor through masked mean pooling, projects this anchor into a compact latent space, and refines the projected representation with residual self-attention and feedforward blocks before producing a single video embedding for the recommender. Due to the redundancy in the frame count of the original benchmark and its overly coarse sampling, we used titles to re-s","authors_text":"Bo Hui, Chao Jiang, Fei Wang, Huiyuan Chen, Kaiyuan Deng, Ruimeng Ye, Xiaolong Ma, Yang Xiao, Zinan Ling","cross_cats":["cs.AI"],"headline":"The Compressed Video Aggregator decouples video content from preference learning by combining frozen VFM embeddings through latent reasoning to create compact representations that improve micro-video recommendation accuracy while cutting训练","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:02:20Z","title":"Compressed Video Aggregator: Content-driven Module for Efficient Micro-Video Recommendation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.08810","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-12T02:05:09.875128Z","id":"20d756c2-38bb-43f8-b910-de65e2c04612","model_set":{"reader":"grok-4.3"},"one_line_summary":"CVA aggregates frozen VFM embeddings via latent reasoning to create compact video embeddings for efficient micro-video recommendation, delivering consistent performance gains and orders-of-magnitude efficiency improvements.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"The Compressed Video Aggregator decouples video content from preference learning by combining frozen VFM embeddings through latent reasoning to create compact representations that improve micro-video recommendation accuracy while cutting训练","strongest_claim":"CVA aggregates frozen VFM embeddings, and uses latent reasoning without cross-attention projection, producing compact video embeddings for recommenders. Experiments show consistent gains with orders-of-magnitude reductions in training time and GPU memory; re-selected frames further enhance performance of all methods.","weakest_assumption":"That embeddings from frozen VFMs plus title-based CLIP key-frame selection capture sufficient video content for preference learning without requiring end-to-end fine-tuning or full video processing."}},"verdict_id":"20d756c2-38bb-43f8-b910-de65e2c04612"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:556fa5a873637bc0b731c8cd3d57ed3df3c4f59461937d8d0d9d9247612d7f8d","target":"record","created_at":"2026-07-30T01:17:00Z","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":"145c87945dfcaf56feb9ff6429aec15b39ce6a3d310d237851f217807a285039","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T09:02:20Z","title_canon_sha256":"cbb19f3fe370e9bbabbc402f222d56c4e3b03826415b17c7d56c430daf44ba42"},"schema_version":"1.0","source":{"id":"2605.08810","kind":"arxiv","version":2}},"canonical_sha256":"0a8cd567d6a01cdf4a9335641ce0760d0e5bd3457bb269e5c53921baf3db4222","receipt":{"builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0a8cd567d6a01cdf4a9335641ce0760d0e5bd3457bb269e5c53921baf3db4222","first_computed_at":"2026-07-30T01:17:00.268214Z","kind":"pith_receipt","last_reissued_at":"2026-07-30T01:17:00.268214Z","receipt_version":"0.3","signature_status":"unsigned_v0"},"source_id":"2605.08810","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:556fa5a873637bc0b731c8cd3d57ed3df3c4f59461937d8d0d9d9247612d7f8d","sha256:c70466493e79ac3062f3a33b852fae5313138b0bb5cb82c699ce0df4bb641e61"],"state_sha256":"8d78042e74860206cd3f7194aa275f4f1507764804a757a32e334a5f23394f56"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"2Ctznisj+qYpEDPPAhrn27Xm1/VrYjdVr2nagdwZ5RYe15COI6wlF24sGhwpumUx5mc1RWtX+3Tp4lv0a6N2DA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-14T08:14:16.514453Z","bundle_sha256":"e8966d0e957cb4a4f6e481e44958336a461d42b2e5888a605b900e1567880b50"}}