{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:GKB7Y23VEDJC5V5JREJCVYO5GV","short_pith_number":"pith:GKB7Y23V","schema_version":"1.0","canonical_sha256":"3283fc6b7520d22ed7a989122ae1dd354b981f7d758e9e48dd9e5fb97a85cc81","source":{"kind":"arxiv","id":"2510.17045","version":2},"attestation_state":"computed","paper":{"title":"Video Reasoning without Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Ankita Nayak, Deepak Sridhar, Harris Teague, Jeya Pradha Jeyaraj, Kartikeya Bhardwaj, Nuno Vasconcelos","submitted_at":"2025-10-19T23:17:13Z","abstract_excerpt":"Video reasoning using Large Multimodal Models (LMMs) relies on costly reinforcement learning (RL) and verbose chain-of-thought, resulting in substantial computational overhead during both training and inference. Moreover, the mechanisms that control the thinking process in these reasoning models are very limited. In this paper, we use the entropy of the model's output distribution as a signal to study and guide reasoning behavior. We discover that high-quality models exhibit a characteristic pattern of micro-exploration and micro-exploitation cycles, followed by a later entropy peak (i.e., lon"},"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":"2510.17045","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-10-19T23:17:13Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"ace8b1b8133e900743d63624db517c80bfce512c934024d9971ecc8baa6b042c","abstract_canon_sha256":"3eef01a6abfe6866057e894754ba3288999fdc12212e4a981d2eb397e350a5a7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T02:04:48.563948Z","signature_b64":"2RtXjuJDlLe8uWQ0bEgLXM6UVp03vVRYil7pHt+73HB43Vl9ieR4mfJ7vkLNDVBEg2CP1plxdF4h/QRHNAgAAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3283fc6b7520d22ed7a989122ae1dd354b981f7d758e9e48dd9e5fb97a85cc81","last_reissued_at":"2026-06-02T02:04:48.563396Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T02:04:48.563396Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Video Reasoning without Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Ankita Nayak, Deepak Sridhar, Harris Teague, Jeya Pradha Jeyaraj, Kartikeya Bhardwaj, Nuno Vasconcelos","submitted_at":"2025-10-19T23:17:13Z","abstract_excerpt":"Video reasoning using Large Multimodal Models (LMMs) relies on costly reinforcement learning (RL) and verbose chain-of-thought, resulting in substantial computational overhead during both training and inference. Moreover, the mechanisms that control the thinking process in these reasoning models are very limited. In this paper, we use the entropy of the model's output distribution as a signal to study and guide reasoning behavior. We discover that high-quality models exhibit a characteristic pattern of micro-exploration and micro-exploitation cycles, followed by a later entropy peak (i.e., lon"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.17045","kind":"arxiv","version":2},"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/2510.17045/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":"2510.17045","created_at":"2026-06-02T02:04:48.563465+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.17045v2","created_at":"2026-06-02T02:04:48.563465+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.17045","created_at":"2026-06-02T02:04:48.563465+00:00"},{"alias_kind":"pith_short_12","alias_value":"GKB7Y23VEDJC","created_at":"2026-06-02T02:04:48.563465+00:00"},{"alias_kind":"pith_short_16","alias_value":"GKB7Y23VEDJC5V5J","created_at":"2026-06-02T02:04:48.563465+00:00"},{"alias_kind":"pith_short_8","alias_value":"GKB7Y23V","created_at":"2026-06-02T02:04:48.563465+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.25642","citing_title":"Prefill-Time Intervention for Mitigating Hallucination in Large Vision-Language Models","ref_index":34,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV","json":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV.json","graph_json":"https://pith.science/api/pith-number/GKB7Y23VEDJC5V5JREJCVYO5GV/graph.json","events_json":"https://pith.science/api/pith-number/GKB7Y23VEDJC5V5JREJCVYO5GV/events.json","paper":"https://pith.science/paper/GKB7Y23V"},"agent_actions":{"view_html":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV","download_json":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV.json","view_paper":"https://pith.science/paper/GKB7Y23V","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.17045&json=true","fetch_graph":"https://pith.science/api/pith-number/GKB7Y23VEDJC5V5JREJCVYO5GV/graph.json","fetch_events":"https://pith.science/api/pith-number/GKB7Y23VEDJC5V5JREJCVYO5GV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV/action/storage_attestation","attest_author":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV/action/author_attestation","sign_citation":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV/action/citation_signature","submit_replication":"https://pith.science/pith/GKB7Y23VEDJC5V5JREJCVYO5GV/action/replication_record"}},"created_at":"2026-06-02T02:04:48.563465+00:00","updated_at":"2026-06-02T02:04:48.563465+00:00"}