{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SPDDBVRNS36U3W2AYSQMMOTDSD","short_pith_number":"pith:SPDDBVRN","schema_version":"1.0","canonical_sha256":"93c630d62d96fd4ddb40c4a0c63a6390d1fe33a219909cc0a01d9a8d3f0f1d3d","source":{"kind":"arxiv","id":"2511.00066","version":4},"attestation_state":"computed","paper":{"title":"Sharpness-Guided Group Relative Policy Optimization via Probability Shaping","license":"http://creativecommons.org/licenses/by/4.0/","headline":"GRPO-SG is a sharpness-guided token-weighted variant of GRPO that downweights high-gradient tokens to stabilize optimization and improve generalization in reinforcement learning with verifiable rewards.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Linh Ngo Van, Trung Le, Tue Le","submitted_at":"2025-10-29T08:07:47Z","abstract_excerpt":"Reinforcement learning with verifiable rewards (RLVR) has become a practical route to improve large language model reasoning, and Group Relative Policy Optimization (GRPO) is a widely used optimizer in this setting. However, RLVR training is typically performed with limited control over generalization. We revisit GRPO through a robustness-based generalization view, where the generalization loss is upper bounded by a combination of the empirical loss and a sharpness surrogate measured by the gradient norm. Building on this perspective, we propose Sharpness-Guided GRPO (GRPO-SG), a simple token-"},"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":"2511.00066","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-10-29T08:07:47Z","cross_cats_sorted":[],"title_canon_sha256":"31f0c3e1f267c875ebb9f805cfd7b65021ac2103745a48b4901fd85355cf81c9","abstract_canon_sha256":"69ae622ee2d31532a890af88e4480e6416911004d68999429d7b8d6f7b2cc7d7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:09:33.672680Z","signature_b64":"zoBZxa80/PVr6s/gMwALgqRurcjDHBdTpf7nCPmGu3wLEg8U2MVv6tLcDbCy/GB/55vjKO//0bgl2iFkLyItCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93c630d62d96fd4ddb40c4a0c63a6390d1fe33a219909cc0a01d9a8d3f0f1d3d","last_reissued_at":"2026-05-18T03:09:33.672129Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:09:33.672129Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sharpness-Guided Group Relative Policy Optimization via Probability Shaping","license":"http://creativecommons.org/licenses/by/4.0/","headline":"GRPO-SG is a sharpness-guided token-weighted variant of GRPO that downweights high-gradient tokens to stabilize optimization and improve generalization in reinforcement learning with verifiable rewards.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Linh Ngo Van, Trung Le, Tue Le","submitted_at":"2025-10-29T08:07:47Z","abstract_excerpt":"Reinforcement learning with verifiable rewards (RLVR) has become a practical route to improve large language model reasoning, and Group Relative Policy Optimization (GRPO) is a widely used optimizer in this setting. However, RLVR training is typically performed with limited control over generalization. We revisit GRPO through a robustness-based generalization view, where the generalization loss is upper bounded by a combination of the empirical loss and a sharpness surrogate measured by the gradient norm. Building on this perspective, we propose Sharpness-Guided GRPO (GRPO-SG), a simple token-"},"claims":{"count":3,"items":[{"kind":"strongest_claim","text":"GRPO-SG, a simple token-weighted variant of GRPO, downweights tokens likely to cause overly large gradients, reducing sharp updates and stabilizing optimization, thereby improving generalization in RLVR.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the generalization loss is upper bounded by a combination of the empirical loss and a sharpness surrogate measured by the gradient norm, and that downweighting high-gradient tokens will reliably reduce this sharpness in the RLVR setting for LLMs.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"GRPO-SG is a sharpness-guided token-weighted variant of GRPO that downweights high-gradient tokens to stabilize optimization and improve generalization in reinforcement learning with verifiable rewards.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"}],"snapshot_sha256":"9b90ce15ea5ceca22a10db72cd910009dc7570f1d5ea55c60b134457067c5ef3"},"source":{"id":"2511.00066","kind":"arxiv","version":4},"verdict":{"id":"90ca2c5f-5ba7-4e36-81e5-c4d09186bcfa","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T03:07:44.282901Z","strongest_claim":"GRPO-SG, a simple token-weighted variant of GRPO, downweights tokens likely to cause overly large gradients, reducing sharp updates and stabilizing optimization, thereby improving generalization in RLVR.","one_line_summary":"GRPO-SG is a sharpness-guided token-weighted variant of GRPO that downweights high-gradient tokens to stabilize optimization and improve generalization in reinforcement learning with verifiable rewards.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the generalization loss is upper bounded by a combination of the empirical loss and a sharpness surrogate measured by the gradient norm, and that downweighting high-gradient tokens will reliably reduce this sharpness in the RLVR setting for LLMs.","pith_extraction_headline":""},"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":"2511.00066","created_at":"2026-05-18T03:09:33.672203+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.00066v4","created_at":"2026-05-18T03:09:33.672203+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.00066","created_at":"2026-05-18T03:09:33.672203+00:00"},{"alias_kind":"pith_short_12","alias_value":"SPDDBVRNS36U","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"SPDDBVRNS36U3W2A","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"SPDDBVRN","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2602.15620","citing_title":"STAPO: Stabilizing Reinforcement Learning for LLMs by Silencing Rare Spurious Tokens","ref_index":10,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD","json":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD.json","graph_json":"https://pith.science/api/pith-number/SPDDBVRNS36U3W2AYSQMMOTDSD/graph.json","events_json":"https://pith.science/api/pith-number/SPDDBVRNS36U3W2AYSQMMOTDSD/events.json","paper":"https://pith.science/paper/SPDDBVRN"},"agent_actions":{"view_html":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD","download_json":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD.json","view_paper":"https://pith.science/paper/SPDDBVRN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.00066&json=true","fetch_graph":"https://pith.science/api/pith-number/SPDDBVRNS36U3W2AYSQMMOTDSD/graph.json","fetch_events":"https://pith.science/api/pith-number/SPDDBVRNS36U3W2AYSQMMOTDSD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD/action/storage_attestation","attest_author":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD/action/author_attestation","sign_citation":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD/action/citation_signature","submit_replication":"https://pith.science/pith/SPDDBVRNS36U3W2AYSQMMOTDSD/action/replication_record"}},"created_at":"2026-05-18T03:09:33.672203+00:00","updated_at":"2026-05-18T03:09:33.672203+00:00"}