{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:Z4MPPAATJJU5JKLULCSB7Y3IK5","short_pith_number":"pith:Z4MPPAAT","schema_version":"1.0","canonical_sha256":"cf18f780134a69d4a97458a41fe36857681db57b71b40dc47d17beb78155712c","source":{"kind":"arxiv","id":"2605.22703","version":1},"attestation_state":"computed","paper":{"title":"Clipping Bottleneck: Stabilizing RLVR via Stochastic Recovery of Near-Boundary Signals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bolin Ding, Chiyu Ma, Guoyin Wang, Haoming Meng, Jinda Lu, Jingren Zhou, Kexin Huang, Li Yuan, Qihui Zhang, Shuo Yang, Yuyang Liu","submitted_at":"2026-05-21T16:45:31Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a central paradigm for scaling LLM reasoning, yet its optimization often suffers from training instability and suboptimal convergence. Through a systematic dissection of clipping-based GRPO-style objectives, we identify the rigid clipping decision induced by hard clipping as a key practical bottleneck in the studied RLVR setups. Specifically, our analysis suggests that informative signals can lie in the near-boundary region just beyond the clipping threshold, and are therefore discarded by the standard hard-clipping rule. Not"},"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":"2605.22703","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T16:45:31Z","cross_cats_sorted":[],"title_canon_sha256":"3164ba46d12f8f82c68c059f88f9cbd049db08abca3ffdf7a36e653109a5aaab","abstract_canon_sha256":"892e4bf6ec75bc0bba3a68de1e7fe922bf119eb7606ddeb2e48cbfe8a00abb90"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T02:04:51.205302Z","signature_b64":"uq6RP8Yc8RO7NRcgszsfX5VN5JikQ4Y0/tImvq+IjBRNaNKMoL7gdxga/yjH61sDUVhntUfCvEwGptLWz+i5DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cf18f780134a69d4a97458a41fe36857681db57b71b40dc47d17beb78155712c","last_reissued_at":"2026-05-22T02:04:51.204462Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T02:04:51.204462Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Clipping Bottleneck: Stabilizing RLVR via Stochastic Recovery of Near-Boundary Signals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bolin Ding, Chiyu Ma, Guoyin Wang, Haoming Meng, Jinda Lu, Jingren Zhou, Kexin Huang, Li Yuan, Qihui Zhang, Shuo Yang, Yuyang Liu","submitted_at":"2026-05-21T16:45:31Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a central paradigm for scaling LLM reasoning, yet its optimization often suffers from training instability and suboptimal convergence. Through a systematic dissection of clipping-based GRPO-style objectives, we identify the rigid clipping decision induced by hard clipping as a key practical bottleneck in the studied RLVR setups. Specifically, our analysis suggests that informative signals can lie in the near-boundary region just beyond the clipping threshold, and are therefore discarded by the standard hard-clipping rule. Not"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22703","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/2605.22703/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.22703","created_at":"2026-05-22T02:04:51.204625+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.22703v1","created_at":"2026-05-22T02:04:51.204625+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22703","created_at":"2026-05-22T02:04:51.204625+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z4MPPAATJJU5","created_at":"2026-05-22T02:04:51.204625+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z4MPPAATJJU5JKLU","created_at":"2026-05-22T02:04:51.204625+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z4MPPAAT","created_at":"2026-05-22T02:04:51.204625+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5","json":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5.json","graph_json":"https://pith.science/api/pith-number/Z4MPPAATJJU5JKLULCSB7Y3IK5/graph.json","events_json":"https://pith.science/api/pith-number/Z4MPPAATJJU5JKLULCSB7Y3IK5/events.json","paper":"https://pith.science/paper/Z4MPPAAT"},"agent_actions":{"view_html":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5","download_json":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5.json","view_paper":"https://pith.science/paper/Z4MPPAAT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.22703&json=true","fetch_graph":"https://pith.science/api/pith-number/Z4MPPAATJJU5JKLULCSB7Y3IK5/graph.json","fetch_events":"https://pith.science/api/pith-number/Z4MPPAATJJU5JKLULCSB7Y3IK5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5/action/storage_attestation","attest_author":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5/action/author_attestation","sign_citation":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5/action/citation_signature","submit_replication":"https://pith.science/pith/Z4MPPAATJJU5JKLULCSB7Y3IK5/action/replication_record"}},"created_at":"2026-05-22T02:04:51.204625+00:00","updated_at":"2026-05-22T02:04:51.204625+00:00"}