{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:M2KQFQTZY6GYBGKAP6QP7DYZUW","short_pith_number":"pith:M2KQFQTZ","schema_version":"1.0","canonical_sha256":"669502c279c78d8099407fa0ff8f19a5a7a416364ed7208ee54d159fe652127e","source":{"kind":"arxiv","id":"2506.20670","version":1},"attestation_state":"computed","paper":{"title":"MMSearch-R1: Incentivizing LMMs to Search","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Reinforcement learning lets multimodal models search the internet only when needed","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bo Li, Bo You, Jinming Wu, Wei Li, Yiding Liu, Zejun Ma, Zihao Deng, Ziwei Liu","submitted_at":"2025-06-25T17:59:42Z","abstract_excerpt":"Robust deployment of large multimodal models (LMMs) in real-world scenarios requires access to external knowledge sources, given the complexity and dynamic nature of real-world information. Existing approaches such as retrieval-augmented generation (RAG) and prompt engineered search agents rely on rigid pipelines, often leading to inefficient or excessive search behaviors. We present MMSearch-R1, the first end-to-end reinforcement learning framework that enables LMMs to perform on-demand, multi-turn search in real-world Internet environments. Our framework integrates both image and text search"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2506.20670","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-06-25T17:59:42Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"8ecc7b053022d2d258e49fd9c9a94ad2866fd3f21cc2e0a15ea9923f28eb1f08","abstract_canon_sha256":"d87506b60c8c6cbcb204c48af3d47e2d19d61c1bbb8a14d3a703a876367c7425"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:47.444920Z","signature_b64":"qekVEaEoh4rFM9FM5LvrDkpfPnAlMrKGR9suWfXA4LF4aQ4eHd1qSYhOp4/Blj0nIWEuc3i2OTcH3zKKL3u8CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"669502c279c78d8099407fa0ff8f19a5a7a416364ed7208ee54d159fe652127e","last_reissued_at":"2026-05-17T23:38:47.444503Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:47.444503Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MMSearch-R1: Incentivizing LMMs to Search","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Reinforcement learning lets multimodal models search the internet only when needed","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bo Li, Bo You, Jinming Wu, Wei Li, Yiding Liu, Zejun Ma, Zihao Deng, Ziwei Liu","submitted_at":"2025-06-25T17:59:42Z","abstract_excerpt":"Robust deployment of large multimodal models (LMMs) in real-world scenarios requires access to external knowledge sources, given the complexity and dynamic nature of real-world information. Existing approaches such as retrieval-augmented generation (RAG) and prompt engineered search agents rely on rigid pipelines, often leading to inefficient or excessive search behaviors. We present MMSearch-R1, the first end-to-end reinforcement learning framework that enables LMMs to perform on-demand, multi-turn search in real-world Internet environments. Our framework integrates both image and text search"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We present MMSearch-R1, the first end-to-end reinforcement learning framework that enables LMMs to perform on-demand, multi-turn search in real-world Internet environments... our model not only outperforms RAG-based baselines of the same model size, but also matches the performance of a larger RAG-based model while reducing search calls by over 30%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The outcome-based reward combined with a search penalty, together with the curated search-balanced dataset, is sufficient to produce efficient on-demand search behavior that generalizes beyond the training distribution.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MMSearch-R1 uses reinforcement learning to train multimodal models for on-demand multi-turn internet search with image and text tools, outperforming same-size RAG baselines and matching larger ones while cutting search calls by over 30%.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Reinforcement learning lets multimodal models search the internet only when needed","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3447c2a81aad5ee45d9eec2cce236f98202eda62d8441b689c80c1450340ee79"},"source":{"id":"2506.20670","kind":"arxiv","version":1},"verdict":{"id":"18f9ea2e-53a7-4c65-865b-759ce7bedc8c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T15:21:56.253302Z","strongest_claim":"We present MMSearch-R1, the first end-to-end reinforcement learning framework that enables LMMs to perform on-demand, multi-turn search in real-world Internet environments... our model not only outperforms RAG-based baselines of the same model size, but also matches the performance of a larger RAG-based model while reducing search calls by over 30%.","one_line_summary":"MMSearch-R1 uses reinforcement learning to train multimodal models for on-demand multi-turn internet search with image and text tools, outperforming same-size RAG baselines and matching larger ones while cutting search calls by over 30%.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The outcome-based reward combined with a search penalty, together with the curated search-balanced dataset, is sufficient to produce efficient on-demand search behavior that generalizes beyond the training distribution.","pith_extraction_headline":"Reinforcement learning lets multimodal models search the internet only when needed"},"references":{"count":94,"sample":[{"doi":"","year":2025,"title":"Open deep search: Democratizing search with open-source reasoning agents","work_id":"467e5b5b-a169-4539-8ddc-bd34e07b0e3a","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Claude 3.5 Sonnet","work_id":"45d58c07-b237-4bb8-ac8d-8c68f4784507","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Self-rag: Learn- ing to retrieve, generate, and critique through self-reflection","work_id":"cd9884d5-53dc-4ba6-98d6-a38b02cf052e","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Mint-1t: Scaling open-source multimodal data by 10x: A multimodal dataset with one trillion tokens","work_id":"9350a215-8bdd-4a04-a217-295413298f83","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","ref_index":5,"cited_arxiv_id":"2502.13923","is_internal_anchor":true}],"resolved_work":94,"snapshot_sha256":"2a80b22b2e8323d5b48d357cc8e67cc82540c731f2a5ec48516d7e6cffe9fca1","internal_anchors":26},"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":"2506.20670","created_at":"2026-05-17T23:38:47.444573+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.20670v1","created_at":"2026-05-17T23:38:47.444573+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.20670","created_at":"2026-05-17T23:38:47.444573+00:00"},{"alias_kind":"pith_short_12","alias_value":"M2KQFQTZY6GY","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"M2KQFQTZY6GYBGKA","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"M2KQFQTZ","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":22,"internal_anchor_count":22,"sample":[{"citing_arxiv_id":"2605.17946","citing_title":"SVFSearch: A Multimodal Knowledge-Intensive Benchmark for Short-Video Frame Search in the Gaming Vertical Domain","ref_index":66,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17946","citing_title":"SVFSearch: A Multimodal Knowledge-Intensive Benchmark for Short-Video Frame Search in the Gaming Vertical Domain","ref_index":66,"is_internal_anchor":true},{"citing_arxiv_id":"2505.22095","citing_title":"Mixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2509.07969","citing_title":"Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2511.05271","citing_title":"DeepEyesV2: Toward Agentic Multimodal Model","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"2602.22683","citing_title":"SUPERGLASSES: Benchmarking Vision Language Models as Intelligent Agents for AI Smart Glasses","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12497","citing_title":"From Web to Pixels: Bringing Agentic Search into Visual Perception","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10832","citing_title":"Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agents","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09934","citing_title":"TRACER: Verifiable Generative Provenance for Multimodal Tool-Using Agents","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09271","citing_title":"Shaping Schema via Language Representation as the Next Frontier for LLM Intelligence Expanding","ref_index":115,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07177","citing_title":"HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2604.20146","citing_title":"SAKE: Self-aware Knowledge Exploitation-Exploration for Grounded Multimodal Named Entity Recognition","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2604.20486","citing_title":"ProMMSearchAgent: A Generalizable Multimodal Search Agent Trained with Process-Oriented Rewards","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2604.19264","citing_title":"DR-MMSearchAgent: Deepening Reasoning in Multimodal Search Agents","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2604.12890","citing_title":"Towards Long-horizon Agentic Multimodal Search","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2604.09508","citing_title":"VISOR: Agentic Visual Retrieval-Augmented Generation via Iterative Search and Over-horizon Reasoning","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07177","citing_title":"HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2604.06777","citing_title":"Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimization","ref_index":64,"is_internal_anchor":true},{"citing_arxiv_id":"2604.08545","citing_title":"Act Wisely: Cultivating Meta-Cognitive Tool Use in Agentic Multimodal Models","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2604.04500","citing_title":"Saliency-R1: Enforcing Interpretable and Faithful Vision-language Reasoning via Saliency-map Alignment Reward","ref_index":75,"is_internal_anchor":true},{"citing_arxiv_id":"2604.14029","citing_title":"POINTS-Seeker: Towards Training a Multimodal Agentic Search Model from Scratch","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02378","citing_title":"Enhancing Multimodal In-Context Learning via Inductive-Deductive Reasoning","ref_index":45,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW","json":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW.json","graph_json":"https://pith.science/api/pith-number/M2KQFQTZY6GYBGKAP6QP7DYZUW/graph.json","events_json":"https://pith.science/api/pith-number/M2KQFQTZY6GYBGKAP6QP7DYZUW/events.json","paper":"https://pith.science/paper/M2KQFQTZ"},"agent_actions":{"view_html":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW","download_json":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW.json","view_paper":"https://pith.science/paper/M2KQFQTZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.20670&json=true","fetch_graph":"https://pith.science/api/pith-number/M2KQFQTZY6GYBGKAP6QP7DYZUW/graph.json","fetch_events":"https://pith.science/api/pith-number/M2KQFQTZY6GYBGKAP6QP7DYZUW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW/action/storage_attestation","attest_author":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW/action/author_attestation","sign_citation":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW/action/citation_signature","submit_replication":"https://pith.science/pith/M2KQFQTZY6GYBGKAP6QP7DYZUW/action/replication_record"}},"created_at":"2026-05-17T23:38:47.444573+00:00","updated_at":"2026-05-17T23:38:47.444573+00:00"}