{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:LP4KFZY3CV7GO4XQ66AT7UOQXR","short_pith_number":"pith:LP4KFZY3","schema_version":"1.0","canonical_sha256":"5bf8a2e71b157e6772f0f7813fd1d0bc420fd06493993702134d7918ca5ee80b","source":{"kind":"arxiv","id":"2606.03892","version":1},"attestation_state":"computed","paper":{"title":"Synthesize and Reward -- Reinforcement Learning for Multi-Step Tool Use in Live Environments","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Asim Munawar, Chulaka Gunasekara, Ibrahim Abdelaziz, Kinjal Basu, Maxwell Crouse, Pavan Kapanipathi, Suneet Katrekar","submitted_at":"2026-06-02T16:52:31Z","abstract_excerpt":"Training LLMs to orchestrate multi-step tool calls is held back by three coupled obstacles: realistic stateful execution environments are costly to build, synthetic training queries are often detached from the server's actual state (so the generated tool calls fail to execute), and recall-based RL rewards incentivize verbose tool-calling patterns. We present PROVE (Programmatic Rewards On Verified Environments), a framework with three contributions: (1) a library of 20 stateful MCP (Model Context Protocol) servers exposing 343 tools, enabling live-execution RL training with session-scoped stat"},"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":"2606.03892","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T16:52:31Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"e9e988780aa74ca57097cd4763ca4c0e416ad0fb71fc246eb23c59202b51aa73","abstract_canon_sha256":"bd2cd98671c91c74f67c1b157034fbe6ced10a20ee7f0fa64129a7285f3722d8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T02:06:06.176180Z","signature_b64":"tTh1Z0JCTUoYJ2Ki8vB4k4rQgU25e+ZBMT6ZaSEbJnHE79NVq3+nebsJ+vT6xNThGkLT8e10vnsigDh6ADJrDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5bf8a2e71b157e6772f0f7813fd1d0bc420fd06493993702134d7918ca5ee80b","last_reissued_at":"2026-06-03T02:06:06.175776Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T02:06:06.175776Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Synthesize and Reward -- Reinforcement Learning for Multi-Step Tool Use in Live Environments","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Asim Munawar, Chulaka Gunasekara, Ibrahim Abdelaziz, Kinjal Basu, Maxwell Crouse, Pavan Kapanipathi, Suneet Katrekar","submitted_at":"2026-06-02T16:52:31Z","abstract_excerpt":"Training LLMs to orchestrate multi-step tool calls is held back by three coupled obstacles: realistic stateful execution environments are costly to build, synthetic training queries are often detached from the server's actual state (so the generated tool calls fail to execute), and recall-based RL rewards incentivize verbose tool-calling patterns. We present PROVE (Programmatic Rewards On Verified Environments), a framework with three contributions: (1) a library of 20 stateful MCP (Model Context Protocol) servers exposing 343 tools, enabling live-execution RL training with session-scoped stat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03892","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/2606.03892/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":"2606.03892","created_at":"2026-06-03T02:06:06.175832+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.03892v1","created_at":"2026-06-03T02:06:06.175832+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03892","created_at":"2026-06-03T02:06:06.175832+00:00"},{"alias_kind":"pith_short_12","alias_value":"LP4KFZY3CV7G","created_at":"2026-06-03T02:06:06.175832+00:00"},{"alias_kind":"pith_short_16","alias_value":"LP4KFZY3CV7GO4XQ","created_at":"2026-06-03T02:06:06.175832+00:00"},{"alias_kind":"pith_short_8","alias_value":"LP4KFZY3","created_at":"2026-06-03T02:06:06.175832+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/LP4KFZY3CV7GO4XQ66AT7UOQXR","json":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR.json","graph_json":"https://pith.science/api/pith-number/LP4KFZY3CV7GO4XQ66AT7UOQXR/graph.json","events_json":"https://pith.science/api/pith-number/LP4KFZY3CV7GO4XQ66AT7UOQXR/events.json","paper":"https://pith.science/paper/LP4KFZY3"},"agent_actions":{"view_html":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR","download_json":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR.json","view_paper":"https://pith.science/paper/LP4KFZY3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.03892&json=true","fetch_graph":"https://pith.science/api/pith-number/LP4KFZY3CV7GO4XQ66AT7UOQXR/graph.json","fetch_events":"https://pith.science/api/pith-number/LP4KFZY3CV7GO4XQ66AT7UOQXR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR/action/storage_attestation","attest_author":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR/action/author_attestation","sign_citation":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR/action/citation_signature","submit_replication":"https://pith.science/pith/LP4KFZY3CV7GO4XQ66AT7UOQXR/action/replication_record"}},"created_at":"2026-06-03T02:06:06.175832+00:00","updated_at":"2026-06-03T02:06:06.175832+00:00"}