{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4LPIUM4K4ZSICKQY6IRO7PWX4Z","short_pith_number":"pith:4LPIUM4K","schema_version":"1.0","canonical_sha256":"e2de8a338ae664812a18f222efbed7e669689ce6d9807d8e607b9d03afae0760","source":{"kind":"arxiv","id":"2412.15606","version":2},"attestation_state":"computed","paper":{"title":"Multi-modal Agent Tuning: Building a VLM-Driven Agent for Efficient Tool Usage","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.AI","authors_text":"Bofei Zhang, Pengxiang Li, Qing Li, Song-Chun Zhu, Tao Yuan, Xiaojian Ma, Yue Fan, Yunde Jia, Yuwei Wu, Zhi Gao","submitted_at":"2024-12-20T07:00:46Z","abstract_excerpt":"The advancement of large language models (LLMs) prompts the development of multi-modal agents, which are used as a controller to call external tools, providing a feasible way to solve practical tasks. In this paper, we propose a multi-modal agent tuning method that automatically generates multi-modal tool-usage data and tunes a vision-language model (VLM) as the controller for powerful tool-usage reasoning. To preserve the data quality, we prompt the GPT-4o mini model to generate queries, files, and trajectories, followed by query-file and trajectory verifiers. Based on the data synthesis pipe"},"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":"2412.15606","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-12-20T07:00:46Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"4258df22b44040723a5bb8c47a5789ff9bc51b62bc6fbc449fb4a20c9e4ae093","abstract_canon_sha256":"6a780f055d99d88f21a3742af55e51ef83c9b5ec17d346d4109e5536c2d6c10b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:08:26.605843Z","signature_b64":"mptRWRE+Q643XjssjspKIsRwWc+CbSheuK4ls60mRBvq3DmVAMc5bYSZb3obot0uSXOOHZfEKgmPqyTFnImqBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e2de8a338ae664812a18f222efbed7e669689ce6d9807d8e607b9d03afae0760","last_reissued_at":"2026-07-05T10:08:26.605350Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:08:26.605350Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-modal Agent Tuning: Building a VLM-Driven Agent for Efficient Tool Usage","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.AI","authors_text":"Bofei Zhang, Pengxiang Li, Qing Li, Song-Chun Zhu, Tao Yuan, Xiaojian Ma, Yue Fan, Yunde Jia, Yuwei Wu, Zhi Gao","submitted_at":"2024-12-20T07:00:46Z","abstract_excerpt":"The advancement of large language models (LLMs) prompts the development of multi-modal agents, which are used as a controller to call external tools, providing a feasible way to solve practical tasks. In this paper, we propose a multi-modal agent tuning method that automatically generates multi-modal tool-usage data and tunes a vision-language model (VLM) as the controller for powerful tool-usage reasoning. To preserve the data quality, we prompt the GPT-4o mini model to generate queries, files, and trajectories, followed by query-file and trajectory verifiers. Based on the data synthesis pipe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.15606","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/2412.15606/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":"2412.15606","created_at":"2026-07-05T10:08:26.605417+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.15606v2","created_at":"2026-07-05T10:08:26.605417+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.15606","created_at":"2026-07-05T10:08:26.605417+00:00"},{"alias_kind":"pith_short_12","alias_value":"4LPIUM4K4ZSI","created_at":"2026-07-05T10:08:26.605417+00:00"},{"alias_kind":"pith_short_16","alias_value":"4LPIUM4K4ZSICKQY","created_at":"2026-07-05T10:08:26.605417+00:00"},{"alias_kind":"pith_short_8","alias_value":"4LPIUM4K","created_at":"2026-07-05T10:08:26.605417+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08497","citing_title":"Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and Editing","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2606.31392","citing_title":"ReGRPO: Reflection-Augmented Policy Optimization for Tool-Using Agents","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2510.02837","citing_title":"Beyond the Final Answer: Evaluating the Reasoning Trajectories of Tool-Augmented Agents","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03631","citing_title":"Single-agent vs. Multi-agents for Automated Video Analysis of On-Screen Collaborative Learning Behaviors","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z","json":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z.json","graph_json":"https://pith.science/api/pith-number/4LPIUM4K4ZSICKQY6IRO7PWX4Z/graph.json","events_json":"https://pith.science/api/pith-number/4LPIUM4K4ZSICKQY6IRO7PWX4Z/events.json","paper":"https://pith.science/paper/4LPIUM4K"},"agent_actions":{"view_html":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z","download_json":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z.json","view_paper":"https://pith.science/paper/4LPIUM4K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.15606&json=true","fetch_graph":"https://pith.science/api/pith-number/4LPIUM4K4ZSICKQY6IRO7PWX4Z/graph.json","fetch_events":"https://pith.science/api/pith-number/4LPIUM4K4ZSICKQY6IRO7PWX4Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z/action/storage_attestation","attest_author":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z/action/author_attestation","sign_citation":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z/action/citation_signature","submit_replication":"https://pith.science/pith/4LPIUM4K4ZSICKQY6IRO7PWX4Z/action/replication_record"}},"created_at":"2026-07-05T10:08:26.605417+00:00","updated_at":"2026-07-05T10:08:26.605417+00:00"}