{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:D5HYECPFDLFK7TLTUIU4TOT6TE","short_pith_number":"pith:D5HYECPF","canonical_record":{"source":{"id":"2607.10848","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-12T17:20:44Z","cross_cats_sorted":[],"title_canon_sha256":"bc1ab58e002173604a716a7e3a3961626b276bf951b6477b923cf6f26fe245c3","abstract_canon_sha256":"36e9516c3bc5abb8013ec7978868f464b6dc5df965e17c3be9938f55f06453ce"},"schema_version":"1.0"},"canonical_sha256":"1f4f8209e51acaafcd73a229c9ba7e9931539f4f63aeb4958d96de4319f6c34c","source":{"kind":"arxiv","id":"2607.10848","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.10848","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"arxiv_version","alias_value":"2607.10848v1","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.10848","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"pith_short_12","alias_value":"D5HYECPFDLFK","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"pith_short_16","alias_value":"D5HYECPFDLFK7TLT","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"pith_short_8","alias_value":"D5HYECPF","created_at":"2026-07-14T01:21:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:D5HYECPFDLFK7TLTUIU4TOT6TE","target":"record","payload":{"canonical_record":{"source":{"id":"2607.10848","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-12T17:20:44Z","cross_cats_sorted":[],"title_canon_sha256":"bc1ab58e002173604a716a7e3a3961626b276bf951b6477b923cf6f26fe245c3","abstract_canon_sha256":"36e9516c3bc5abb8013ec7978868f464b6dc5df965e17c3be9938f55f06453ce"},"schema_version":"1.0"},"canonical_sha256":"1f4f8209e51acaafcd73a229c9ba7e9931539f4f63aeb4958d96de4319f6c34c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T01:21:43.400579Z","signature_b64":"AMc73wTLtlxkfL0KhJT/GSI6+4j2DDjLhbAfoZzqQ0woMK+vozrPZQ8vugX93lNyBQOTSwg/D7/znH2QRSUSCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1f4f8209e51acaafcd73a229c9ba7e9931539f4f63aeb4958d96de4319f6c34c","last_reissued_at":"2026-07-14T01:21:43.399699Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T01:21:43.399699Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2607.10848","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-14T01:21:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"J9bLJU8OpOpwpjdPGBj2mm7lsTfL3I/tprfW5fJfbzmu5qsqHZZP6s8zB7KIkHaaxtXgJL1wSwVp1jQJ+UhrCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T09:04:13.751893Z"},"content_sha256":"32d178850f592fb8108108c1a4be9025d6fcc1b4ccb7cf06d6d7a33e1408486a","schema_version":"1.0","event_id":"sha256:32d178850f592fb8108108c1a4be9025d6fcc1b4ccb7cf06d6d7a33e1408486a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:D5HYECPFDLFK7TLTUIU4TOT6TE","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Predictive Divergence Masks for LLM RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bowen Ping, Haonan Wang, Jiaqi Tang, Jiarui Yao, Penghui Qi, Tianyu Pang, Xiangxin Zhou","submitted_at":"2026-07-12T17:20:44Z","abstract_excerpt":"Reinforcement learning for large language models (LLMs) typically relies on trust-region masks to stabilize off-policy updates. The dominant PPO-style approach uses the sampled-token importance ratio for two criteria: a proximity criterion, which asks whether the policy has moved too far from the behavior policy, and a direction criterion, which asks whether the update pushes it farther away. Recent work DPPO improves the proximity criterion by replacing PPO's ratio-based test with a probability divergence between the behavior and training policies. However, its direction criterion is still in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.10848","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/2607.10848/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-14T01:21:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qdkY+jGhZQtgu6pe7ba5nCCsRVxQPS+Yu0vgHZZoDscLrq62ssdbZT3mgN1qXyoKAWjEuDHwOohQFnw+0XZBBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T09:04:13.752813Z"},"content_sha256":"4681f8ec63ab6e907ca6f6d2716a5c6bc5f85fc2619eceb0600aac1d5946e9d7","schema_version":"1.0","event_id":"sha256:4681f8ec63ab6e907ca6f6d2716a5c6bc5f85fc2619eceb0600aac1d5946e9d7"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/D5HYECPFDLFK7TLTUIU4TOT6TE/bundle.json","state_url":"https://pith.science/pith/D5HYECPFDLFK7TLTUIU4TOT6TE/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/D5HYECPFDLFK7TLTUIU4TOT6TE/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-03T09:04:13Z","links":{"resolver":"https://pith.science/pith/D5HYECPFDLFK7TLTUIU4TOT6TE","bundle":"https://pith.science/pith/D5HYECPFDLFK7TLTUIU4TOT6TE/bundle.json","state":"https://pith.science/pith/D5HYECPFDLFK7TLTUIU4TOT6TE/state.json","well_known_bundle":"https://pith.science/.well-known/pith/D5HYECPFDLFK7TLTUIU4TOT6TE/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:D5HYECPFDLFK7TLTUIU4TOT6TE","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":"36e9516c3bc5abb8013ec7978868f464b6dc5df965e17c3be9938f55f06453ce","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-12T17:20:44Z","title_canon_sha256":"bc1ab58e002173604a716a7e3a3961626b276bf951b6477b923cf6f26fe245c3"},"schema_version":"1.0","source":{"id":"2607.10848","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.10848","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"arxiv_version","alias_value":"2607.10848v1","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.10848","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"pith_short_12","alias_value":"D5HYECPFDLFK","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"pith_short_16","alias_value":"D5HYECPFDLFK7TLT","created_at":"2026-07-14T01:21:43Z"},{"alias_kind":"pith_short_8","alias_value":"D5HYECPF","created_at":"2026-07-14T01:21:43Z"}],"graph_snapshots":[{"event_id":"sha256:4681f8ec63ab6e907ca6f6d2716a5c6bc5f85fc2619eceb0600aac1d5946e9d7","target":"graph","created_at":"2026-07-14T01:21:43Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2607.10848/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning for large language models (LLMs) typically relies on trust-region masks to stabilize off-policy updates. The dominant PPO-style approach uses the sampled-token importance ratio for two criteria: a proximity criterion, which asks whether the policy has moved too far from the behavior policy, and a direction criterion, which asks whether the update pushes it farther away. Recent work DPPO improves the proximity criterion by replacing PPO's ratio-based test with a probability divergence between the behavior and training policies. However, its direction criterion is still in","authors_text":"Bowen Ping, Haonan Wang, Jiaqi Tang, Jiarui Yao, Penghui Qi, Tianyu Pang, Xiangxin Zhou","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-12T17:20:44Z","title":"Predictive Divergence Masks for LLM RL"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.10848","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:32d178850f592fb8108108c1a4be9025d6fcc1b4ccb7cf06d6d7a33e1408486a","target":"record","created_at":"2026-07-14T01:21:43Z","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":"36e9516c3bc5abb8013ec7978868f464b6dc5df965e17c3be9938f55f06453ce","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-12T17:20:44Z","title_canon_sha256":"bc1ab58e002173604a716a7e3a3961626b276bf951b6477b923cf6f26fe245c3"},"schema_version":"1.0","source":{"id":"2607.10848","kind":"arxiv","version":1}},"canonical_sha256":"1f4f8209e51acaafcd73a229c9ba7e9931539f4f63aeb4958d96de4319f6c34c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"1f4f8209e51acaafcd73a229c9ba7e9931539f4f63aeb4958d96de4319f6c34c","first_computed_at":"2026-07-14T01:21:43.399699Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-14T01:21:43.399699Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"AMc73wTLtlxkfL0KhJT/GSI6+4j2DDjLhbAfoZzqQ0woMK+vozrPZQ8vugX93lNyBQOTSwg/D7/znH2QRSUSCw==","signature_status":"signed_v1","signed_at":"2026-07-14T01:21:43.400579Z","signed_message":"canonical_sha256_bytes"},"source_id":"2607.10848","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:32d178850f592fb8108108c1a4be9025d6fcc1b4ccb7cf06d6d7a33e1408486a","sha256:4681f8ec63ab6e907ca6f6d2716a5c6bc5f85fc2619eceb0600aac1d5946e9d7"],"state_sha256":"95e025b27cd4d6e2dd4eaa2146c8ef7cc413357a50213e32331a19dd3dc41eea"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"dbI7TwoIF+83AzEAFKPj1xclGSZIW9+r6+Jdr6zOtfmEohb+l6SgjK5MXUG43YSbO/803bZCcld0opWcHVRUDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T09:04:13.768340Z","bundle_sha256":"30106bc33b5b898162e48818b3102264c060487c95288cbb65de6a8ef0b2222f"}}