{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RYIPWZWZFDHVPJXBTBHN7NEHKF","short_pith_number":"pith:RYIPWZWZ","schema_version":"1.0","canonical_sha256":"8e10fb66d928cf57a6e1984edfb487515c3368e3f7ec5c2cb55beed61a45b67d","source":{"kind":"arxiv","id":"2606.05800","version":1},"attestation_state":"computed","paper":{"title":"SALT: When More Rollouts Don't Help in Group-Based Policy Optimization and How to Make Them Matter","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chaoqun Sun, Chenyu Wang, Dongying Kong, JianXiang Xiang, Jinpeng Zhang, Lianrui Li, MiniWell Tsao, Powei Chang, Yukang Gao","submitted_at":"2026-06-04T07:29:43Z","abstract_excerpt":"Reinforcement learning with verifiable rewards (RLVR) often adopts GRPO-style group-relative updates, sampling multiple rollouts per prompt to construct normalized learning signals. However, merely increasing the number of rollouts does not reliably strengthen learning: under GRPO-style group normalization, per-rollout policy-gradient features can concentrate into a low-rank, signed geometry, causing substantial cancellation during aggregation and weakening the effective update. We address this failure mode with SALT, a Subspace-Adaptive geometry pLug-in componenT that uses sample-wise gradien"},"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":"2606.05800","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T07:29:43Z","cross_cats_sorted":[],"title_canon_sha256":"e602ce86fb22619cd2bb2015f71816355849b774e6de50a690b54a184c8e7835","abstract_canon_sha256":"5f21879db8810a7a800e62370ac48d766bcaba7bc50967c63759a2446c401044"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:15:03.774101Z","signature_b64":"iDfZdl8Wr8fYkSRxhn3OxTbyrVSDHUkXonBD35ngo6BaaCdXv7kVwPI2KMlbFCppz9JOBctegYyoWfqZW7KJCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8e10fb66d928cf57a6e1984edfb487515c3368e3f7ec5c2cb55beed61a45b67d","last_reissued_at":"2026-06-05T01:15:03.773492Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:15:03.773492Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SALT: When More Rollouts Don't Help in Group-Based Policy Optimization and How to Make Them Matter","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chaoqun Sun, Chenyu Wang, Dongying Kong, JianXiang Xiang, Jinpeng Zhang, Lianrui Li, MiniWell Tsao, Powei Chang, Yukang Gao","submitted_at":"2026-06-04T07:29:43Z","abstract_excerpt":"Reinforcement learning with verifiable rewards (RLVR) often adopts GRPO-style group-relative updates, sampling multiple rollouts per prompt to construct normalized learning signals. However, merely increasing the number of rollouts does not reliably strengthen learning: under GRPO-style group normalization, per-rollout policy-gradient features can concentrate into a low-rank, signed geometry, causing substantial cancellation during aggregation and weakening the effective update. We address this failure mode with SALT, a Subspace-Adaptive geometry pLug-in componenT that uses sample-wise gradien"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.05800","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/2606.05800/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":"2606.05800","created_at":"2026-06-05T01:15:03.773567+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.05800v1","created_at":"2026-06-05T01:15:03.773567+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.05800","created_at":"2026-06-05T01:15:03.773567+00:00"},{"alias_kind":"pith_short_12","alias_value":"RYIPWZWZFDHV","created_at":"2026-06-05T01:15:03.773567+00:00"},{"alias_kind":"pith_short_16","alias_value":"RYIPWZWZFDHVPJXB","created_at":"2026-06-05T01:15:03.773567+00:00"},{"alias_kind":"pith_short_8","alias_value":"RYIPWZWZ","created_at":"2026-06-05T01:15:03.773567+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/RYIPWZWZFDHVPJXBTBHN7NEHKF","json":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF.json","graph_json":"https://pith.science/api/pith-number/RYIPWZWZFDHVPJXBTBHN7NEHKF/graph.json","events_json":"https://pith.science/api/pith-number/RYIPWZWZFDHVPJXBTBHN7NEHKF/events.json","paper":"https://pith.science/paper/RYIPWZWZ"},"agent_actions":{"view_html":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF","download_json":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF.json","view_paper":"https://pith.science/paper/RYIPWZWZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.05800&json=true","fetch_graph":"https://pith.science/api/pith-number/RYIPWZWZFDHVPJXBTBHN7NEHKF/graph.json","fetch_events":"https://pith.science/api/pith-number/RYIPWZWZFDHVPJXBTBHN7NEHKF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF/action/storage_attestation","attest_author":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF/action/author_attestation","sign_citation":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF/action/citation_signature","submit_replication":"https://pith.science/pith/RYIPWZWZFDHVPJXBTBHN7NEHKF/action/replication_record"}},"created_at":"2026-06-05T01:15:03.773567+00:00","updated_at":"2026-06-05T01:15:03.773567+00:00"}