{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AW4GZPZDBURNRDFL5EUU5DFHEQ","short_pith_number":"pith:AW4GZPZD","schema_version":"1.0","canonical_sha256":"05b86cbf230d22d88cabe9294e8ca7240c858f588c11acc5f6fcf2298a935034","source":{"kind":"arxiv","id":"2608.01717","version":1},"attestation_state":"computed","paper":{"title":"Beyond On-Policy Exploration: Integrating External Policy Rollouts for Reinforcement Learning in Diffusion Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jimyeong Kim, Jungmin Ko, Wonjong Rhee, Wonseok Lee","submitted_at":"2026-08-03T05:28:37Z","abstract_excerpt":"Recent reinforcement learning methods for diffusion large language models (dLLMs) commonly rely on on-policy rollouts generated by the target dLLM itself. When successful on-policy rollouts are scarce, however, on-policy training may receive little positive reward and make only limited progress. To mitigate this problem, we explore incorporating higher-reward rollouts generated by a stronger external policy alongside on-policy rollouts from the target dLLM. However, directly incorporating these external rollouts introduces two practical challenges: differences in rollout length and instability"},"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":"2608.01717","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-08-03T05:28:37Z","cross_cats_sorted":[],"title_canon_sha256":"f08a73c5429f858d374ee835a9450e323642e6f24c141d3c238ba8498168e9c5","abstract_canon_sha256":"eeaf74cc0473697ab02bb972c39bc20c079784cc67278cf5baf753aacac0aded"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T02:06:36.998708Z","signature_b64":"tD8BMXHbw+FRflrWzl7kuTn+kfsPCnS954ckx5yIBekrfwSEjT4175vqNuC3CPyIjCNhPWuNzhLKZeaObz7DBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"05b86cbf230d22d88cabe9294e8ca7240c858f588c11acc5f6fcf2298a935034","last_reissued_at":"2026-08-04T02:06:36.997085Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T02:06:36.997085Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond On-Policy Exploration: Integrating External Policy Rollouts for Reinforcement Learning in Diffusion Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jimyeong Kim, Jungmin Ko, Wonjong Rhee, Wonseok Lee","submitted_at":"2026-08-03T05:28:37Z","abstract_excerpt":"Recent reinforcement learning methods for diffusion large language models (dLLMs) commonly rely on on-policy rollouts generated by the target dLLM itself. When successful on-policy rollouts are scarce, however, on-policy training may receive little positive reward and make only limited progress. To mitigate this problem, we explore incorporating higher-reward rollouts generated by a stronger external policy alongside on-policy rollouts from the target dLLM. However, directly incorporating these external rollouts introduces two practical challenges: differences in rollout length and instability"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.01717","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/2608.01717/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":"2608.01717","created_at":"2026-08-04T02:06:36.998626+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.01717v1","created_at":"2026-08-04T02:06:36.998626+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.01717","created_at":"2026-08-04T02:06:36.998626+00:00"},{"alias_kind":"pith_short_12","alias_value":"AW4GZPZDBURN","created_at":"2026-08-04T02:06:36.998626+00:00"},{"alias_kind":"pith_short_16","alias_value":"AW4GZPZDBURNRDFL","created_at":"2026-08-04T02:06:36.998626+00:00"},{"alias_kind":"pith_short_8","alias_value":"AW4GZPZD","created_at":"2026-08-04T02:06:36.998626+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/AW4GZPZDBURNRDFL5EUU5DFHEQ","json":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ.json","graph_json":"https://pith.science/api/pith-number/AW4GZPZDBURNRDFL5EUU5DFHEQ/graph.json","events_json":"https://pith.science/api/pith-number/AW4GZPZDBURNRDFL5EUU5DFHEQ/events.json","paper":"https://pith.science/paper/AW4GZPZD"},"agent_actions":{"view_html":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ","download_json":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ.json","view_paper":"https://pith.science/paper/AW4GZPZD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.01717&json=true","fetch_graph":"https://pith.science/api/pith-number/AW4GZPZDBURNRDFL5EUU5DFHEQ/graph.json","fetch_events":"https://pith.science/api/pith-number/AW4GZPZDBURNRDFL5EUU5DFHEQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ/action/storage_attestation","attest_author":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ/action/author_attestation","sign_citation":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ/action/citation_signature","submit_replication":"https://pith.science/pith/AW4GZPZDBURNRDFL5EUU5DFHEQ/action/replication_record"}},"created_at":"2026-08-04T02:06:36.998626+00:00","updated_at":"2026-08-04T02:06:36.998626+00:00"}