{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IQGIMFN2WN372YFSB3KPHI72SK","short_pith_number":"pith:IQGIMFN2","schema_version":"1.0","canonical_sha256":"440c8615bab377fd60b20ed4f3a3fa928c8ae3de34bf15539c03d3a5154387a6","source":{"kind":"arxiv","id":"2506.14821","version":3},"attestation_state":"computed","paper":{"title":"Reinforcing VLMs to Use Tools for Detailed Visual Reasoning Under Resource Constraints","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Bowen Zhao, Leo Dirac, Paulina Varshavskaya, Sunil Kumar","submitted_at":"2025-06-10T20:11:44Z","abstract_excerpt":"Despite tremendous recent advances in large model reasoning ability, vision-language models (VLMs) still struggle with detailed visual reasoning, especially when compute resources are limited. To address this challenge, we draw inspiration from methods like Deepseek-r1 for VLMs and train smaller-scale models with Group Relative Policy Optimization (GRPO) to use external tools such as zoom. The greatest benefit is obtained with a combination of GRPO learning, a simple reward structure, a simplified tool-calling interface, allocating additional tokens to the result of the tool call, and a traini"},"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":"2506.14821","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-10T20:11:44Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"03d84ab0a2d362b345b34d83308e9ba082caf1994ad02db5254a7eddd636def9","abstract_canon_sha256":"d05252881397b6f611c8db3e349b6578698a5fb54deab99a2fab751ba75a1305"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:48:26.771279Z","signature_b64":"Iu2cFViy2A6s9Uge8nCFQnQASDhWufL4LFqM0mQW4s88kYMpzgtHeLnjpyCon0ePNJCAbejEtigYRWOR2ORLBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"440c8615bab377fd60b20ed4f3a3fa928c8ae3de34bf15539c03d3a5154387a6","last_reissued_at":"2026-07-05T11:48:26.770799Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:48:26.770799Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reinforcing VLMs to Use Tools for Detailed Visual Reasoning Under Resource Constraints","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Bowen Zhao, Leo Dirac, Paulina Varshavskaya, Sunil Kumar","submitted_at":"2025-06-10T20:11:44Z","abstract_excerpt":"Despite tremendous recent advances in large model reasoning ability, vision-language models (VLMs) still struggle with detailed visual reasoning, especially when compute resources are limited. To address this challenge, we draw inspiration from methods like Deepseek-r1 for VLMs and train smaller-scale models with Group Relative Policy Optimization (GRPO) to use external tools such as zoom. The greatest benefit is obtained with a combination of GRPO learning, a simple reward structure, a simplified tool-calling interface, allocating additional tokens to the result of the tool call, and a traini"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.14821","kind":"arxiv","version":3},"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/2506.14821/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":"2506.14821","created_at":"2026-07-05T11:48:26.770852+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.14821v3","created_at":"2026-07-05T11:48:26.770852+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.14821","created_at":"2026-07-05T11:48:26.770852+00:00"},{"alias_kind":"pith_short_12","alias_value":"IQGIMFN2WN37","created_at":"2026-07-05T11:48:26.770852+00:00"},{"alias_kind":"pith_short_16","alias_value":"IQGIMFN2WN372YFS","created_at":"2026-07-05T11:48:26.770852+00:00"},{"alias_kind":"pith_short_8","alias_value":"IQGIMFN2","created_at":"2026-07-05T11:48:26.770852+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26196","citing_title":"From Structure to Synergy: A Survey of Vision-Language Perception Paradigm Evolution in Multimodal Large Language Models","ref_index":208,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20980","citing_title":"Robusto-2: Benchmarking Humans & VLMs for Autonomous Driving in Lima & New York City","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19945","citing_title":"Visual Reasoning through Tool-supervised Reinforcement Learning","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK","json":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK.json","graph_json":"https://pith.science/api/pith-number/IQGIMFN2WN372YFSB3KPHI72SK/graph.json","events_json":"https://pith.science/api/pith-number/IQGIMFN2WN372YFSB3KPHI72SK/events.json","paper":"https://pith.science/paper/IQGIMFN2"},"agent_actions":{"view_html":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK","download_json":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK.json","view_paper":"https://pith.science/paper/IQGIMFN2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.14821&json=true","fetch_graph":"https://pith.science/api/pith-number/IQGIMFN2WN372YFSB3KPHI72SK/graph.json","fetch_events":"https://pith.science/api/pith-number/IQGIMFN2WN372YFSB3KPHI72SK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK/action/storage_attestation","attest_author":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK/action/author_attestation","sign_citation":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK/action/citation_signature","submit_replication":"https://pith.science/pith/IQGIMFN2WN372YFSB3KPHI72SK/action/replication_record"}},"created_at":"2026-07-05T11:48:26.770852+00:00","updated_at":"2026-07-05T11:48:26.770852+00:00"}