{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:3HU2VDJGKHSIYOYYLTZCFSMJ3L","short_pith_number":"pith:3HU2VDJG","canonical_record":{"source":{"id":"2606.04560","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T07:47:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8cea3f8f7ca4b0a0351b6983b36289e16ef9bd4b2ab3f0df0ec343a5dd82a84f","abstract_canon_sha256":"6c8222607e3b5480a1a209bb404dcd2dc91548e6f9de4f843678ee39126bbc8b"},"schema_version":"1.0"},"canonical_sha256":"d9e9aa8d2651e48c3b185cf222c989dac8b21504c86aa77d91fe6027a7b95870","source":{"kind":"arxiv","id":"2606.04560","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.04560","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"arxiv_version","alias_value":"2606.04560v1","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04560","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"pith_short_12","alias_value":"3HU2VDJGKHSI","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"pith_short_16","alias_value":"3HU2VDJGKHSIYOYY","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"pith_short_8","alias_value":"3HU2VDJG","created_at":"2026-06-04T01:09:18Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:3HU2VDJGKHSIYOYYLTZCFSMJ3L","target":"record","payload":{"canonical_record":{"source":{"id":"2606.04560","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T07:47:47Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8cea3f8f7ca4b0a0351b6983b36289e16ef9bd4b2ab3f0df0ec343a5dd82a84f","abstract_canon_sha256":"6c8222607e3b5480a1a209bb404dcd2dc91548e6f9de4f843678ee39126bbc8b"},"schema_version":"1.0"},"canonical_sha256":"d9e9aa8d2651e48c3b185cf222c989dac8b21504c86aa77d91fe6027a7b95870","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:09:18.159461Z","signature_b64":"y57mEwzcwJ/bR6fFCnOaIWMwuixtkD1R8qnyxkKc0xYSK1PLdEQc/OKFro0O/Ni9M2l3xeYXGEwKDjLmWEj9DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d9e9aa8d2651e48c3b185cf222c989dac8b21504c86aa77d91fe6027a7b95870","last_reissued_at":"2026-06-04T01:09:18.158629Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:09:18.158629Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2606.04560","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":"2026-06-04T01:09:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XScdsbwIz7GBPs/ZxC6y4Wui0x5/w5qgMf3orRJu1NC6gWlm32yOUjPh0okCBK8gImLsq4IbBOab58ANhUdFDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-09T19:53:24.092401Z"},"content_sha256":"8df93d158761139ec62bcac20943834e5ec69a113ed1460831645f233a6d5b5a","schema_version":"1.0","event_id":"sha256:8df93d158761139ec62bcac20943834e5ec69a113ed1460831645f233a6d5b5a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:3HU2VDJGKHSIYOYYLTZCFSMJ3L","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Rollout-Level Advantage-Prioritized Experience Replay for GRPO","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Gyeongtae Yoo, Ik-hwan Kim, Sanghyeok Park, Soohyuk Jang, Sungroh Yoon","submitted_at":"2026-06-03T07:47:47Z","abstract_excerpt":"Reinforcement learning from verifiable rewards with GRPO is a standard approach for post-training reasoning LLMs. It remains sample inefficient. Each rollout is used for a single gradient update and then discarded. Naive replay is not well suited in this setting because LLM policies drift quickly per gradient step. Stored rollouts therefore become stale and can destabilize training. We propose a rollout-level replay buffer for GRPO that stores and samples individual rollouts rather than whole groups. The buffer bounds staleness through age eviction. Any rollout older than tau_max training step"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04560","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/2606.04560/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":"2026-06-04T01:09:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6aJ0aSIdcM/xk4ygjFdHm1/CkyYZwq93ueSDODNM9DpsbUJQto/cEuN2ph5p6ckz6lTd9C2tFsquNeKGaYLjBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-09T19:53:24.093182Z"},"content_sha256":"8f7c9f769983a2897d958d13c3da33fdb6d03b5fe717ed2dbca712c0d8f00c0f","schema_version":"1.0","event_id":"sha256:8f7c9f769983a2897d958d13c3da33fdb6d03b5fe717ed2dbca712c0d8f00c0f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L/bundle.json","state_url":"https://pith.science/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L/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":"2026-06-09T19:53:24Z","links":{"resolver":"https://pith.science/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L","bundle":"https://pith.science/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L/bundle.json","state":"https://pith.science/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3HU2VDJGKHSIYOYYLTZCFSMJ3L/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:3HU2VDJGKHSIYOYYLTZCFSMJ3L","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":"6c8222607e3b5480a1a209bb404dcd2dc91548e6f9de4f843678ee39126bbc8b","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T07:47:47Z","title_canon_sha256":"8cea3f8f7ca4b0a0351b6983b36289e16ef9bd4b2ab3f0df0ec343a5dd82a84f"},"schema_version":"1.0","source":{"id":"2606.04560","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.04560","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"arxiv_version","alias_value":"2606.04560v1","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04560","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"pith_short_12","alias_value":"3HU2VDJGKHSI","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"pith_short_16","alias_value":"3HU2VDJGKHSIYOYY","created_at":"2026-06-04T01:09:18Z"},{"alias_kind":"pith_short_8","alias_value":"3HU2VDJG","created_at":"2026-06-04T01:09:18Z"}],"graph_snapshots":[{"event_id":"sha256:8f7c9f769983a2897d958d13c3da33fdb6d03b5fe717ed2dbca712c0d8f00c0f","target":"graph","created_at":"2026-06-04T01:09:18Z","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/2606.04560/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning from verifiable rewards with GRPO is a standard approach for post-training reasoning LLMs. It remains sample inefficient. Each rollout is used for a single gradient update and then discarded. Naive replay is not well suited in this setting because LLM policies drift quickly per gradient step. Stored rollouts therefore become stale and can destabilize training. We propose a rollout-level replay buffer for GRPO that stores and samples individual rollouts rather than whole groups. The buffer bounds staleness through age eviction. Any rollout older than tau_max training step","authors_text":"Gyeongtae Yoo, Ik-hwan Kim, Sanghyeok Park, Soohyuk Jang, Sungroh Yoon","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T07:47:47Z","title":"Rollout-Level Advantage-Prioritized Experience Replay for GRPO"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04560","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":"sha256:8df93d158761139ec62bcac20943834e5ec69a113ed1460831645f233a6d5b5a","target":"record","created_at":"2026-06-04T01:09:18Z","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":"6c8222607e3b5480a1a209bb404dcd2dc91548e6f9de4f843678ee39126bbc8b","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T07:47:47Z","title_canon_sha256":"8cea3f8f7ca4b0a0351b6983b36289e16ef9bd4b2ab3f0df0ec343a5dd82a84f"},"schema_version":"1.0","source":{"id":"2606.04560","kind":"arxiv","version":1}},"canonical_sha256":"d9e9aa8d2651e48c3b185cf222c989dac8b21504c86aa77d91fe6027a7b95870","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d9e9aa8d2651e48c3b185cf222c989dac8b21504c86aa77d91fe6027a7b95870","first_computed_at":"2026-06-04T01:09:18.158629Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-04T01:09:18.158629Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"y57mEwzcwJ/bR6fFCnOaIWMwuixtkD1R8qnyxkKc0xYSK1PLdEQc/OKFro0O/Ni9M2l3xeYXGEwKDjLmWEj9DQ==","signature_status":"signed_v1","signed_at":"2026-06-04T01:09:18.159461Z","signed_message":"canonical_sha256_bytes"},"source_id":"2606.04560","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8df93d158761139ec62bcac20943834e5ec69a113ed1460831645f233a6d5b5a","sha256:8f7c9f769983a2897d958d13c3da33fdb6d03b5fe717ed2dbca712c0d8f00c0f"],"state_sha256":"7db75b3f06f11fade40485127aad9fa9b937f7ae2fcb95100be66dbfc2668845"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vmg7nGzepA/Rh96+hYdoXgsG8+QMxZOanGeIUDGacxPhAVMzQRieqbLqtOZHx9Llpxagh6jufUH43R+MpkUaBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-09T19:53:24.098039Z","bundle_sha256":"28cd6dd497c02844c1d22d20fc300748e143a2e6196c6604b14496605aaa5e25"}}