{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2BLPPV5ZTN3Y3S7AMDX3MIDVXN","short_pith_number":"pith:2BLPPV5Z","schema_version":"1.0","canonical_sha256":"d056f7d7b99b778dcbe060efb62075bb570a8da2fdd663dbf96ba17e16a6c06a","source":{"kind":"arxiv","id":"2607.14171","version":1},"attestation_state":"computed","paper":{"title":"Branching Policy Optimization: Sandbox-Native Language Agent Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Bowei He, Xiaokun Zhang, Xue Liu, Yankai Chen","submitted_at":"2026-07-15T09:37:36Z","abstract_excerpt":"Reinforcement learning has emerged as the dominant paradigm for training large language model (LLM) agents that interact with executable sandboxes. State-of-the-art algorithms such as PPO, RLOO, and GRPO inherit their rollout topology from RLHF: for each prompt, N independent trajectories are sampled from the initial state, and an advantage is computed by subtracting a group baseline. This design ignores a defining property of agent sandboxes. They are deterministic, snapshottable, and resumable from any intermediate state. We argue that this property enables a fundamentally different rollout "},"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":"2607.14171","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-07-15T09:37:36Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"5e2f1bfd381708fcd099ed38c6d0fb69183c17fbf44563125eb408bfdbd7002f","abstract_canon_sha256":"35b2a9f9de71a1263aff23d9ecd2a0873c1be01b9670df887409275e36e84025"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T00:20:55.083410Z","signature_b64":"zA1TwQxarSDJA6oGPKF7t1On0IH1Nyw0yrSFPTGw7Dgo/Ws+iLKN/Y2QNysbCjb2tHh5OmYb989Cni4JZg53Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d056f7d7b99b778dcbe060efb62075bb570a8da2fdd663dbf96ba17e16a6c06a","last_reissued_at":"2026-07-17T00:20:55.082581Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T00:20:55.082581Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Branching Policy Optimization: Sandbox-Native Language Agent Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Bowei He, Xiaokun Zhang, Xue Liu, Yankai Chen","submitted_at":"2026-07-15T09:37:36Z","abstract_excerpt":"Reinforcement learning has emerged as the dominant paradigm for training large language model (LLM) agents that interact with executable sandboxes. State-of-the-art algorithms such as PPO, RLOO, and GRPO inherit their rollout topology from RLHF: for each prompt, N independent trajectories are sampled from the initial state, and an advantage is computed by subtracting a group baseline. This design ignores a defining property of agent sandboxes. They are deterministic, snapshottable, and resumable from any intermediate state. We argue that this property enables a fundamentally different rollout "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14171","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/2607.14171/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":"2607.14171","created_at":"2026-07-17T00:20:55.082997+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14171v1","created_at":"2026-07-17T00:20:55.082997+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14171","created_at":"2026-07-17T00:20:55.082997+00:00"},{"alias_kind":"pith_short_12","alias_value":"2BLPPV5ZTN3Y","created_at":"2026-07-17T00:20:55.082997+00:00"},{"alias_kind":"pith_short_16","alias_value":"2BLPPV5ZTN3Y3S7A","created_at":"2026-07-17T00:20:55.082997+00:00"},{"alias_kind":"pith_short_8","alias_value":"2BLPPV5Z","created_at":"2026-07-17T00:20:55.082997+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/2BLPPV5ZTN3Y3S7AMDX3MIDVXN","json":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN.json","graph_json":"https://pith.science/api/pith-number/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/graph.json","events_json":"https://pith.science/api/pith-number/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/events.json","paper":"https://pith.science/paper/2BLPPV5Z"},"agent_actions":{"view_html":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN","download_json":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN.json","view_paper":"https://pith.science/paper/2BLPPV5Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14171&json=true","fetch_graph":"https://pith.science/api/pith-number/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/graph.json","fetch_events":"https://pith.science/api/pith-number/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/action/storage_attestation","attest_author":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/action/author_attestation","sign_citation":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/action/citation_signature","submit_replication":"https://pith.science/pith/2BLPPV5ZTN3Y3S7AMDX3MIDVXN/action/replication_record"}},"created_at":"2026-07-17T00:20:55.082997+00:00","updated_at":"2026-07-17T00:20:55.082997+00:00"}