{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WLWEA3WS7RUYXF24ARAY7S4ALQ","short_pith_number":"pith:WLWEA3WS","schema_version":"1.0","canonical_sha256":"b2ec406ed2fc698b975c04418fcb805c3e5f818bf713f2f34ed4e04fd659b699","source":{"kind":"arxiv","id":"2502.12081","version":1},"attestation_state":"computed","paper":{"title":"Unhackable Temporal Rewarding for Scalable Video MLLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"En Yu, Haoran Wei, Jianjian Sun, Jingyu Wang, Kangheng Lin, Liang Zhao, Wenbing Tao, Xiangyu Zhang, Yana Wei, Zheng Ge, Zining Zhu","submitted_at":"2025-02-17T17:55:55Z","abstract_excerpt":"In the pursuit of superior video-processing MLLMs, we have encountered a perplexing paradox: the \"anti-scaling law\", where more data and larger models lead to worse performance. This study unmasks the culprit: \"temporal hacking\", a phenomenon where models shortcut by fixating on select frames, missing the full video narrative. In this work, we systematically establish a comprehensive theory of temporal hacking, defining it from a reinforcement learning perspective, introducing the Temporal Perplexity (TPL) score to assess this misalignment, and proposing the Unhackable Temporal Rewarding (UTR)"},"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":"2502.12081","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-02-17T17:55:55Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2a3d1396079fcb2005f066879ae15b821b8db389753f445c64de883b5da0b7f5","abstract_canon_sha256":"762cdd36b6156ad7bc569563a6e4464b340f9038db03b6dd5ccf4968f85bac79"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:38.673593Z","signature_b64":"ezUm4m2NDknjAg/B3wLhge38iD0xHmp14MLn4bVym/OA3VoJJbM0vGIAnHYUiNnk4POkBSozGZdDMPR4u/O5CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b2ec406ed2fc698b975c04418fcb805c3e5f818bf713f2f34ed4e04fd659b699","last_reissued_at":"2026-07-05T10:15:38.673043Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:38.673043Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unhackable Temporal Rewarding for Scalable Video MLLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"En Yu, Haoran Wei, Jianjian Sun, Jingyu Wang, Kangheng Lin, Liang Zhao, Wenbing Tao, Xiangyu Zhang, Yana Wei, Zheng Ge, Zining Zhu","submitted_at":"2025-02-17T17:55:55Z","abstract_excerpt":"In the pursuit of superior video-processing MLLMs, we have encountered a perplexing paradox: the \"anti-scaling law\", where more data and larger models lead to worse performance. This study unmasks the culprit: \"temporal hacking\", a phenomenon where models shortcut by fixating on select frames, missing the full video narrative. In this work, we systematically establish a comprehensive theory of temporal hacking, defining it from a reinforcement learning perspective, introducing the Temporal Perplexity (TPL) score to assess this misalignment, and proposing the Unhackable Temporal Rewarding (UTR)"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12081","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/2502.12081/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":"2502.12081","created_at":"2026-07-05T10:15:38.673099+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.12081v1","created_at":"2026-07-05T10:15:38.673099+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12081","created_at":"2026-07-05T10:15:38.673099+00:00"},{"alias_kind":"pith_short_12","alias_value":"WLWEA3WS7RUY","created_at":"2026-07-05T10:15:38.673099+00:00"},{"alias_kind":"pith_short_16","alias_value":"WLWEA3WS7RUYXF24","created_at":"2026-07-05T10:15:38.673099+00:00"},{"alias_kind":"pith_short_8","alias_value":"WLWEA3WS","created_at":"2026-07-05T10:15:38.673099+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19927","citing_title":"CARE: Competence-Aware Reward Shaping for Adaptive Reasoning Length in Video-MLLMs","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18441","citing_title":"Reasoning as Intersection: Consensus-Frame Alignment for Visual Focus in Video-MLLMs","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21988","citing_title":"Learning Spatiotemporal Sensitivity in Video LLMs via Counterfactual Reinforcement Learning","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2505.20715","citing_title":"MUSEG: Reinforcing Video Temporal Understanding via Timestamp-Aware Multi-Segment Grounding","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2511.11113","citing_title":"VIDEOP2R: Video Understanding from Perception to Reasoning","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2503.21776","citing_title":"Video-R1: Reinforcing Video Reasoning in MLLMs","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01402","citing_title":"Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01324","citing_title":"Beyond Perceptual Shortcuts: Causal-Inspired Debiasing Optimization for Generalizable Video Reasoning in Lightweight MLLMs","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01402","citing_title":"Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ","json":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ.json","graph_json":"https://pith.science/api/pith-number/WLWEA3WS7RUYXF24ARAY7S4ALQ/graph.json","events_json":"https://pith.science/api/pith-number/WLWEA3WS7RUYXF24ARAY7S4ALQ/events.json","paper":"https://pith.science/paper/WLWEA3WS"},"agent_actions":{"view_html":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ","download_json":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ.json","view_paper":"https://pith.science/paper/WLWEA3WS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.12081&json=true","fetch_graph":"https://pith.science/api/pith-number/WLWEA3WS7RUYXF24ARAY7S4ALQ/graph.json","fetch_events":"https://pith.science/api/pith-number/WLWEA3WS7RUYXF24ARAY7S4ALQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ/action/storage_attestation","attest_author":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ/action/author_attestation","sign_citation":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ/action/citation_signature","submit_replication":"https://pith.science/pith/WLWEA3WS7RUYXF24ARAY7S4ALQ/action/replication_record"}},"created_at":"2026-07-05T10:15:38.673099+00:00","updated_at":"2026-07-05T10:15:38.673099+00:00"}