{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2DI52UWCA2IZ2PZVFARPFYGUBC","short_pith_number":"pith:2DI52UWC","schema_version":"1.0","canonical_sha256":"d0d1dd52c206919d3f352822f2e0d408b3d74bf0ff62c39e63f595929c50533c","source":{"kind":"arxiv","id":"2605.28774","version":1},"attestation_state":"computed","paper":{"title":"Agent Explorative Policy Optimization for Multimodal Agentic Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Byung-Kwan Lee, Minki Kang, Pavlo Molchanov, Ryo Hachiuma, Shizhe Diao, Sung Ju Hwang, Yu-Chiang Frank Wang","submitted_at":"2026-05-27T17:36:39Z","abstract_excerpt":"Vision-language models with extended reasoning succeed on complex problems, but many real-world problems require external tools that internal reasoning alone often cannot resolve. Agentic reasoning therefore interleaves two behaviors with a structural asymmetry: thinking (the self-contained default) and tool use (a high-variance auxiliary acting). We refer to this asymmetry as the Thinking-Acting Gap. Under standard RL recipes like GRPO, the gap manifests as two diagnostic symptoms during training: tool use is attempted on only ~30% of rollouts, and when attempted, the tool-using rollouts with"},"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":"2605.28774","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-27T17:36:39Z","cross_cats_sorted":[],"title_canon_sha256":"c3b6fb0d9606c83f3274d84276dad068b836abdcd9d662b6bf694e12093ee968","abstract_canon_sha256":"291fa37a2175ed64bc48384f65eaed6b1765a7349f91d59402517b96e69e2517"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T02:05:02.713209Z","signature_b64":"SAXdCJHaeRO7VRT8Eg1g9ZFlqdWR7YMidlEoXGB77BLFHVKUmXHJjlhV5rw2247NprCHDJRtzkcNg63yiOqVDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d0d1dd52c206919d3f352822f2e0d408b3d74bf0ff62c39e63f595929c50533c","last_reissued_at":"2026-05-28T02:05:02.712770Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T02:05:02.712770Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Agent Explorative Policy Optimization for Multimodal Agentic Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Byung-Kwan Lee, Minki Kang, Pavlo Molchanov, Ryo Hachiuma, Shizhe Diao, Sung Ju Hwang, Yu-Chiang Frank Wang","submitted_at":"2026-05-27T17:36:39Z","abstract_excerpt":"Vision-language models with extended reasoning succeed on complex problems, but many real-world problems require external tools that internal reasoning alone often cannot resolve. Agentic reasoning therefore interleaves two behaviors with a structural asymmetry: thinking (the self-contained default) and tool use (a high-variance auxiliary acting). We refer to this asymmetry as the Thinking-Acting Gap. Under standard RL recipes like GRPO, the gap manifests as two diagnostic symptoms during training: tool use is attempted on only ~30% of rollouts, and when attempted, the tool-using rollouts with"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28774","kind":"arxiv","version":1},"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/2605.28774/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":"2605.28774","created_at":"2026-05-28T02:05:02.712840+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.28774v1","created_at":"2026-05-28T02:05:02.712840+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28774","created_at":"2026-05-28T02:05:02.712840+00:00"},{"alias_kind":"pith_short_12","alias_value":"2DI52UWCA2IZ","created_at":"2026-05-28T02:05:02.712840+00:00"},{"alias_kind":"pith_short_16","alias_value":"2DI52UWCA2IZ2PZV","created_at":"2026-05-28T02:05:02.712840+00:00"},{"alias_kind":"pith_short_8","alias_value":"2DI52UWC","created_at":"2026-05-28T02:05:02.712840+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC","json":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC.json","graph_json":"https://pith.science/api/pith-number/2DI52UWCA2IZ2PZVFARPFYGUBC/graph.json","events_json":"https://pith.science/api/pith-number/2DI52UWCA2IZ2PZVFARPFYGUBC/events.json","paper":"https://pith.science/paper/2DI52UWC"},"agent_actions":{"view_html":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC","download_json":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC.json","view_paper":"https://pith.science/paper/2DI52UWC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.28774&json=true","fetch_graph":"https://pith.science/api/pith-number/2DI52UWCA2IZ2PZVFARPFYGUBC/graph.json","fetch_events":"https://pith.science/api/pith-number/2DI52UWCA2IZ2PZVFARPFYGUBC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC/action/storage_attestation","attest_author":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC/action/author_attestation","sign_citation":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC/action/citation_signature","submit_replication":"https://pith.science/pith/2DI52UWCA2IZ2PZVFARPFYGUBC/action/replication_record"}},"created_at":"2026-05-28T02:05:02.712840+00:00","updated_at":"2026-05-28T02:05:02.712840+00:00"}