{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:LWK7VBFA7CBQDKVKVILWGL6A57","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"071a7288cc5839152eda0dfaf391407930a80f73c3b18801f55879ce8b95ff49","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-27T11:50:26Z","title_canon_sha256":"49042bc26768a5363f48d7758d734ea584e7de7c8e22c1743c4ce050639a39ed"},"schema_version":"1.0","source":{"id":"2604.24357","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.24357","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"arxiv_version","alias_value":"2604.24357v2","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.24357","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_12","alias_value":"LWK7VBFA7CBQ","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_16","alias_value":"LWK7VBFA7CBQDKVK","created_at":"2026-06-19T16:09:58Z"},{"alias_kind":"pith_short_8","alias_value":"LWK7VBFA","created_at":"2026-06-19T16:09:58Z"}],"graph_snapshots":[{"event_id":"sha256:d6ecfbe4307946641e2881711595bf0d169a50e56489a0f8c1690308eef2c598","target":"graph","created_at":"2026-06-19T16:09:58Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"DPRM improves over confidence-based baselines in pretraining, post-training, test-time scaling, and single-cell masked diffusion, with particularly strong gains on harder reasoning subsets. In protein, molecular generation and DNA design, the effect is more multi-objective: ordering-aware variants significantly improve selected structural or fragment-constrained metrics while not uniformly dominating the host baseline on every quality metric."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the online bucketized controller tracks the exact DPRM score at empirical-Bernstein rates and that tractable optimization assumptions hold to deliver sample-complexity advantage over random and confidence-only ordering."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DPRM introduces a Doob h-transform process reward module as a plug-in for token ordering in diffusion language models, with convergence proofs and empirical gains over confidence baselines especially on hard reasoning and scientific design tasks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"DPRM introduces a plug-in module that shifts token ordering in diffusion language models from confidence rules to Doob h-transform process reward guidance."}],"snapshot_sha256":"8bdfce2e6b9985791c5c4583a9f00ec0138e6b23b391ff5da20c406da8fd670b"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-21T06:42:00.638736Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T22:14:02.501214Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.24357/integrity.json","findings":[],"snapshot_sha256":"7a145bc4fe3da201ba25241462a0d3e175f8d5c2eddb830d7f625367efce8c8c","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Diffusion language models generate without a fixed left-to-right order, leaving token ordering as a central algorithmic choice. Existing systems mainly use random masking or confidence-driven ordering, which respectively suffer from train--test mismatch and myopic exploration. We introduce DPRM (Doob -transform Process Reward Model), a plug-in token-ordering module that keeps the host architecture, denoising objective and supervision unchanged, and modifies only the ordering policy. DPRM starts from confidence-driven ordering and gradually shifts to process-reward-guided ordering through onlin","authors_text":"Andi Han, Atsushi Nitanda, Dake Bu, Hau-San Wong, Qingfu Zhang, Taiji Suzuki, Wei Huang","cross_cats":["cs.AI"],"headline":"DPRM introduces a plug-in module that shifts token ordering in diffusion language models from confidence rules to Doob h-transform process reward guidance.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-27T11:50:26Z","title":"DPRM: A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.24357","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-08T04:09:31.083964Z","id":"7dcb89f5-76c2-4738-988f-538c44569db7","model_set":{"reader":"grok-4.3"},"one_line_summary":"DPRM introduces a Doob h-transform process reward module as a plug-in for token ordering in diffusion language models, with convergence proofs and empirical gains over confidence baselines especially on hard reasoning and scientific design tasks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"DPRM introduces a plug-in module that shifts token ordering in diffusion language models from confidence rules to Doob h-transform process reward guidance.","strongest_claim":"DPRM improves over confidence-based baselines in pretraining, post-training, test-time scaling, and single-cell masked diffusion, with particularly strong gains on harder reasoning subsets. In protein, molecular generation and DNA design, the effect is more multi-objective: ordering-aware variants significantly improve selected structural or fragment-constrained metrics while not uniformly dominating the host baseline on every quality metric.","weakest_assumption":"That the online bucketized controller tracks the exact DPRM score at empirical-Bernstein rates and that tractable optimization assumptions hold to deliver sample-complexity advantage over random and confidence-only ordering."}},"verdict_id":"7dcb89f5-76c2-4738-988f-538c44569db7"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:4c35d5ace825fc1ac70fc3ab39921406b21514a34379e80119f921aa058e966e","target":"record","created_at":"2026-06-19T16:09:58Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"071a7288cc5839152eda0dfaf391407930a80f73c3b18801f55879ce8b95ff49","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-27T11:50:26Z","title_canon_sha256":"49042bc26768a5363f48d7758d734ea584e7de7c8e22c1743c4ce050639a39ed"},"schema_version":"1.0","source":{"id":"2604.24357","kind":"arxiv","version":2}},"canonical_sha256":"5d95fa84a0f88301aaaaaa17632fc0eff9c44a5eb7e213828739691fad4c1246","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"5d95fa84a0f88301aaaaaa17632fc0eff9c44a5eb7e213828739691fad4c1246","first_computed_at":"2026-06-19T16:09:58.559402Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-19T16:09:58.559402Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"QjsEj+f0hvqTGdygxkXaleIheLHbyFFjJAN2V0b18NsLJ+ju/I/mq42iOucbYEvSVnluKF6tqOcyWgwYgWI7Bg==","signature_status":"signed_v1","signed_at":"2026-06-19T16:09:58.559831Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.24357","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:4c35d5ace825fc1ac70fc3ab39921406b21514a34379e80119f921aa058e966e","sha256:d6ecfbe4307946641e2881711595bf0d169a50e56489a0f8c1690308eef2c598"],"state_sha256":"b08d2f685402620abe46ddc8dec5472a44318be03900e65d755f67a5ad46cfcd"}