{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QBUSGPPT3Y2ZDMAY6PYAV7P3SA","short_pith_number":"pith:QBUSGPPT","schema_version":"1.0","canonical_sha256":"8069233df3de3591b018f3f00afdfb9006498e0b9f2c5856210f3f2812bfb9a1","source":{"kind":"arxiv","id":"2504.00883","version":2},"attestation_state":"computed","paper":{"title":"Improved Visual-Spatial Reasoning via R1-Zero-Like Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Haonan Lu, Qingsong Xie, Yanhao Zhang, Zhenyi Liao, Zhenyu Yang, Zhijie Deng, Zijian Kong","submitted_at":"2025-04-01T15:11:11Z","abstract_excerpt":"Increasing attention has been placed on improving the reasoning capacities of multi-modal large language models (MLLMs). As the cornerstone for AI agents that function in the physical realm, video-based visual-spatial intelligence (VSI) emerges as one of the most pivotal reasoning capabilities of MLLMs. This work conducts a first, in-depth study on improving the visual-spatial reasoning of MLLMs via R1-Zero-like training. Technically, we first identify that the visual-spatial reasoning capacities of small- to medium-sized Qwen2-VL models cannot be activated via Chain of Thought (CoT) prompts. "},"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":"2504.00883","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-04-01T15:11:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9878dce7df006b1ee104b9bf9936d21918c87ea0f4548e71e2878ef9174aad2a","abstract_canon_sha256":"9f5c68723bc17cd4b62187ccd408a8ed5ebfae74a1b119d9bfe1204bfdd7cc81"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:49:16.245742Z","signature_b64":"UFjvsqFPHSCZ51/Hz4aQf4a2zVYAIfF3RteXF7LF1FN+oH6x3ym/vne7neTuu+javs8H8NZhWR3tFIPJprDaAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8069233df3de3591b018f3f00afdfb9006498e0b9f2c5856210f3f2812bfb9a1","last_reissued_at":"2026-07-05T10:49:16.245262Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:49:16.245262Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improved Visual-Spatial Reasoning via R1-Zero-Like Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Haonan Lu, Qingsong Xie, Yanhao Zhang, Zhenyi Liao, Zhenyu Yang, Zhijie Deng, Zijian Kong","submitted_at":"2025-04-01T15:11:11Z","abstract_excerpt":"Increasing attention has been placed on improving the reasoning capacities of multi-modal large language models (MLLMs). As the cornerstone for AI agents that function in the physical realm, video-based visual-spatial intelligence (VSI) emerges as one of the most pivotal reasoning capabilities of MLLMs. This work conducts a first, in-depth study on improving the visual-spatial reasoning of MLLMs via R1-Zero-like training. Technically, we first identify that the visual-spatial reasoning capacities of small- to medium-sized Qwen2-VL models cannot be activated via Chain of Thought (CoT) prompts. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.00883","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/2504.00883/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":"2504.00883","created_at":"2026-07-05T10:49:16.245322+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.00883v2","created_at":"2026-07-05T10:49:16.245322+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.00883","created_at":"2026-07-05T10:49:16.245322+00:00"},{"alias_kind":"pith_short_12","alias_value":"QBUSGPPT3Y2Z","created_at":"2026-07-05T10:49:16.245322+00:00"},{"alias_kind":"pith_short_16","alias_value":"QBUSGPPT3Y2ZDMAY","created_at":"2026-07-05T10:49:16.245322+00:00"},{"alias_kind":"pith_short_8","alias_value":"QBUSGPPT","created_at":"2026-07-05T10:49:16.245322+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01784","citing_title":"SpaceEra++: A Unified Framework Towards 3D Spatial Reasoning in Video","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11683","citing_title":"Reason, Then Re-reason: Cross-view Revisiting Improves Spatial Reasoning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07433","citing_title":"Watch, Remember, Reason: Human-View Video Understanding with MLLMs","ref_index":219,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01247","citing_title":"Where to Look: Can Foundation Models Reach a Target Viewpoint Through Active Exploration?","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2603.14184","citing_title":"Deeper Thought, Weaker Aim: Understanding and Mitigating Perceptual Impairment during Reasoning in Multimodal Large Language Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20165","citing_title":"CaMo: Camera Motion Grounded Evaluation and Training for Vision-Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2507.00748","citing_title":"Improving the Reasoning of Multi-Image Grounding in MLLMs via Reinforcement Learning","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2505.21374","citing_title":"Video-Holmes: Can MLLM Think Like Holmes for Complex Video Reasoning?","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2503.13377","citing_title":"Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03318","citing_title":"EgoMind: Activating Spatial Cognition through Linguistic Reasoning in MLLMs","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20705","citing_title":"SSL-R1: Self-Supervised Visual Reinforcement Post-Training for Multimodal Large Language Models","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12908","citing_title":"Robotic Manipulation is Vision-to-Geometry Mapping ($f(v) \\rightarrow G$): Vision-Geometry Backbones over Language and Video Models","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06725","citing_title":"Enhancing MLLM Spatial Understanding via Active 3D Scene Exploration for Multi-Perspective Reasoning","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA","json":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA.json","graph_json":"https://pith.science/api/pith-number/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/graph.json","events_json":"https://pith.science/api/pith-number/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/events.json","paper":"https://pith.science/paper/QBUSGPPT"},"agent_actions":{"view_html":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA","download_json":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA.json","view_paper":"https://pith.science/paper/QBUSGPPT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.00883&json=true","fetch_graph":"https://pith.science/api/pith-number/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/graph.json","fetch_events":"https://pith.science/api/pith-number/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/action/storage_attestation","attest_author":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/action/author_attestation","sign_citation":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/action/citation_signature","submit_replication":"https://pith.science/pith/QBUSGPPT3Y2ZDMAY6PYAV7P3SA/action/replication_record"}},"created_at":"2026-07-05T10:49:16.245322+00:00","updated_at":"2026-07-05T10:49:16.245322+00:00"}