{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:OFT3HZQ7BVZKEDQE5LVMKDHIOY","short_pith_number":"pith:OFT3HZQ7","canonical_record":{"source":{"id":"2606.06058","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T11:58:59Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"385b174d97b67f97953c411c616d24772f92d5ff5ae8d68b8bfdb6c065de6fd0","abstract_canon_sha256":"30b141d9574ac9f637ed0bc5d879aa9a5fe7fe9930444e7b3326427693830873"},"schema_version":"1.0"},"canonical_sha256":"7167b3e61f0d72a20e04eaeac50ce876153f4c348d46c5c4fa60197fca09efed","source":{"kind":"arxiv","id":"2606.06058","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.06058","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"arxiv_version","alias_value":"2606.06058v1","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.06058","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"pith_short_12","alias_value":"OFT3HZQ7BVZK","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"pith_short_16","alias_value":"OFT3HZQ7BVZKEDQE","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"pith_short_8","alias_value":"OFT3HZQ7","created_at":"2026-06-05T01:15:31Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:OFT3HZQ7BVZKEDQE5LVMKDHIOY","target":"record","payload":{"canonical_record":{"source":{"id":"2606.06058","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T11:58:59Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"385b174d97b67f97953c411c616d24772f92d5ff5ae8d68b8bfdb6c065de6fd0","abstract_canon_sha256":"30b141d9574ac9f637ed0bc5d879aa9a5fe7fe9930444e7b3326427693830873"},"schema_version":"1.0"},"canonical_sha256":"7167b3e61f0d72a20e04eaeac50ce876153f4c348d46c5c4fa60197fca09efed","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:15:31.761205Z","signature_b64":"c8DWSaYTcH06RwpyrolLqG6YTdYifBx7z0bYPenlxrFGCIUgZ1jsRNy2YQsNSmAJhRUzGuhOaVxnVhta1yqPCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7167b3e61f0d72a20e04eaeac50ce876153f4c348d46c5c4fa60197fca09efed","last_reissued_at":"2026-06-05T01:15:31.760806Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:15:31.760806Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2606.06058","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":"2026-06-05T01:15:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"S+hVtyMwhcnMIz9BTNUfPXW88/amU9LpifJsuQ8D+irEgRbZ5KLEuA2YFS4mVFTrgO1BIzpSnS7l4AH1pJo/DA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-25T11:56:30.445250Z"},"content_sha256":"75b39d96e0b49a2655695befd544d6b39c789112861dccdb10651f7999ae8449","schema_version":"1.0","event_id":"sha256:75b39d96e0b49a2655695befd544d6b39c789112861dccdb10651f7999ae8449"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:OFT3HZQ7BVZKEDQE5LVMKDHIOY","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Heshaam Faili, Mohammad Javad Dousti, Mohammad Mahdi Salmani-Zarchi, Zahra Rahimi","submitted_at":"2026-06-04T11:58:59Z","abstract_excerpt":"Reinforcement learning with verifiable rewards is ideal for multi-constraint instruction following, yet standard group-relative policy optimization (GRPO) becomes unstable under discrete, low-dispersion rewards, where within-group reward distributions are frequently homogeneous. We identify and formalize three pathologies of z-score group normalization in this regime: low-variance amplification, mean-centering blindness, and zero-variance collapse. To address them, we propose MDP-GRPO, which stabilizes learning through (1) multi-temperature sampling to increase reward dispersion, (2) dual-anch"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.06058","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/2606.06058/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":"2026-06-05T01:15:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HA7mxxu6skvRbZzK+XlV23UKoCnmiEI0iKkmuxf78eUd7qIEBVrtbfI1mrGks/gbbDG9Y3Zsh4AUkJQ7rsJVDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-25T11:56:30.445675Z"},"content_sha256":"08ebf3bfee2f6e07e06ae9096a1cf424bf0c26d2cd1ae786a7dae27d8a4315fb","schema_version":"1.0","event_id":"sha256:08ebf3bfee2f6e07e06ae9096a1cf424bf0c26d2cd1ae786a7dae27d8a4315fb"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY/bundle.json","state_url":"https://pith.science/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY/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":"2026-06-25T11:56:30Z","links":{"resolver":"https://pith.science/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY","bundle":"https://pith.science/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY/bundle.json","state":"https://pith.science/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY/state.json","well_known_bundle":"https://pith.science/.well-known/pith/OFT3HZQ7BVZKEDQE5LVMKDHIOY/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:OFT3HZQ7BVZKEDQE5LVMKDHIOY","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":"30b141d9574ac9f637ed0bc5d879aa9a5fe7fe9930444e7b3326427693830873","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T11:58:59Z","title_canon_sha256":"385b174d97b67f97953c411c616d24772f92d5ff5ae8d68b8bfdb6c065de6fd0"},"schema_version":"1.0","source":{"id":"2606.06058","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.06058","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"arxiv_version","alias_value":"2606.06058v1","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.06058","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"pith_short_12","alias_value":"OFT3HZQ7BVZK","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"pith_short_16","alias_value":"OFT3HZQ7BVZKEDQE","created_at":"2026-06-05T01:15:31Z"},{"alias_kind":"pith_short_8","alias_value":"OFT3HZQ7","created_at":"2026-06-05T01:15:31Z"}],"graph_snapshots":[{"event_id":"sha256:08ebf3bfee2f6e07e06ae9096a1cf424bf0c26d2cd1ae786a7dae27d8a4315fb","target":"graph","created_at":"2026-06-05T01:15:31Z","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/2606.06058/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning with verifiable rewards is ideal for multi-constraint instruction following, yet standard group-relative policy optimization (GRPO) becomes unstable under discrete, low-dispersion rewards, where within-group reward distributions are frequently homogeneous. We identify and formalize three pathologies of z-score group normalization in this regime: low-variance amplification, mean-centering blindness, and zero-variance collapse. To address them, we propose MDP-GRPO, which stabilizes learning through (1) multi-temperature sampling to increase reward dispersion, (2) dual-anch","authors_text":"Heshaam Faili, Mohammad Javad Dousti, Mohammad Mahdi Salmani-Zarchi, Zahra Rahimi","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T11:58:59Z","title":"MDP-GRPO: Stabilized Group Relative Policy Optimization for Multi-Constraint Instruction Following"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.06058","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":"sha256:75b39d96e0b49a2655695befd544d6b39c789112861dccdb10651f7999ae8449","target":"record","created_at":"2026-06-05T01:15:31Z","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":"30b141d9574ac9f637ed0bc5d879aa9a5fe7fe9930444e7b3326427693830873","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T11:58:59Z","title_canon_sha256":"385b174d97b67f97953c411c616d24772f92d5ff5ae8d68b8bfdb6c065de6fd0"},"schema_version":"1.0","source":{"id":"2606.06058","kind":"arxiv","version":1}},"canonical_sha256":"7167b3e61f0d72a20e04eaeac50ce876153f4c348d46c5c4fa60197fca09efed","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"7167b3e61f0d72a20e04eaeac50ce876153f4c348d46c5c4fa60197fca09efed","first_computed_at":"2026-06-05T01:15:31.760806Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-05T01:15:31.760806Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"c8DWSaYTcH06RwpyrolLqG6YTdYifBx7z0bYPenlxrFGCIUgZ1jsRNy2YQsNSmAJhRUzGuhOaVxnVhta1yqPCg==","signature_status":"signed_v1","signed_at":"2026-06-05T01:15:31.761205Z","signed_message":"canonical_sha256_bytes"},"source_id":"2606.06058","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:75b39d96e0b49a2655695befd544d6b39c789112861dccdb10651f7999ae8449","sha256:08ebf3bfee2f6e07e06ae9096a1cf424bf0c26d2cd1ae786a7dae27d8a4315fb"],"state_sha256":"b3c4b43ceec45ffed1d9adb51070db9a70e1d836509ce76acc9c248f02e98273"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8SRo6z1gaFATpKnHjiHs88bDRwN1YwWO8GL6IowvL5ymizrn1I9MBZH/g90PnOB94faZi7/qqmio5VRy0iVUCw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-25T11:56:30.447711Z","bundle_sha256":"6ef740d7b295cf1ea28821fcb255dbdc7a80f5fd313c4751f679694308152787"}}