{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YQMUAHGAEMTDSGPZNMAB4SJUYL","short_pith_number":"pith:YQMUAHGA","schema_version":"1.0","canonical_sha256":"c419401cc023263919f96b001e4934c2ef804490c66fb43ff6ff1640d12d09fd","source":{"kind":"arxiv","id":"2506.02096","version":1},"attestation_state":"computed","paper":{"title":"SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesis","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Hang Yan, Jinjie Ni, Michael Qizhe Shieh, Xiangyan Liu, Zichen Liu, Zijian Wu","submitted_at":"2025-06-02T17:45:16Z","abstract_excerpt":"Vision-language models (VLMs) trained via reinforcement learning with verifiable reward (RLVR) have shown notable progress in scaling test-time compute effectively. In this work, we investigate how synthesized RL data can further improve RLVR. To this end, we propose \\textbf{SynthRL}-a scalable and guaranteed pipeline for automatic data scaling in reasoning-oriented RL training. SynthRL comprises three key stages: (1) selecting seed questions with appropriate distribution, (2) augmenting them into more challenging variants while preserving the original answers, and (3) a guaranteed verificatio"},"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":"2506.02096","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-02T17:45:16Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"96f1199076d33bfa8c2415b831a0db03d5b10205a08ccfea5c82a0436af90f3d","abstract_canon_sha256":"64435704e286deb5c89dc8ec292a9c9026b42a5e3b369fb3bfa833b472dca408"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:14:32.340669Z","signature_b64":"4aoNc1Uc4GspGlRFeleKoUOD0WP8TtFjPUPXuTWqmRY6yUcJYJJFUAaObfTrUScfe9uasS1bRUGqRNXN3bkOBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c419401cc023263919f96b001e4934c2ef804490c66fb43ff6ff1640d12d09fd","last_reissued_at":"2026-07-05T11:14:32.340207Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:14:32.340207Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesis","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Hang Yan, Jinjie Ni, Michael Qizhe Shieh, Xiangyan Liu, Zichen Liu, Zijian Wu","submitted_at":"2025-06-02T17:45:16Z","abstract_excerpt":"Vision-language models (VLMs) trained via reinforcement learning with verifiable reward (RLVR) have shown notable progress in scaling test-time compute effectively. In this work, we investigate how synthesized RL data can further improve RLVR. To this end, we propose \\textbf{SynthRL}-a scalable and guaranteed pipeline for automatic data scaling in reasoning-oriented RL training. SynthRL comprises three key stages: (1) selecting seed questions with appropriate distribution, (2) augmenting them into more challenging variants while preserving the original answers, and (3) a guaranteed verificatio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.02096","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/2506.02096/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":"2506.02096","created_at":"2026-07-05T11:14:32.340269+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.02096v1","created_at":"2026-07-05T11:14:32.340269+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.02096","created_at":"2026-07-05T11:14:32.340269+00:00"},{"alias_kind":"pith_short_12","alias_value":"YQMUAHGAEMTD","created_at":"2026-07-05T11:14:32.340269+00:00"},{"alias_kind":"pith_short_16","alias_value":"YQMUAHGAEMTDSGPZ","created_at":"2026-07-05T11:14:32.340269+00:00"},{"alias_kind":"pith_short_8","alias_value":"YQMUAHGA","created_at":"2026-07-05T11:14:32.340269+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2601.04809","citing_title":"SCALER:Synthetic Scalable Adaptive Learning Environment for Reasoning","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00371","citing_title":"GaMMA: Towards Joint Global-Temporal Music Understanding in Large Multimodal Models","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02913","citing_title":"Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning","ref_index":142,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL","json":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL.json","graph_json":"https://pith.science/api/pith-number/YQMUAHGAEMTDSGPZNMAB4SJUYL/graph.json","events_json":"https://pith.science/api/pith-number/YQMUAHGAEMTDSGPZNMAB4SJUYL/events.json","paper":"https://pith.science/paper/YQMUAHGA"},"agent_actions":{"view_html":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL","download_json":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL.json","view_paper":"https://pith.science/paper/YQMUAHGA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.02096&json=true","fetch_graph":"https://pith.science/api/pith-number/YQMUAHGAEMTDSGPZNMAB4SJUYL/graph.json","fetch_events":"https://pith.science/api/pith-number/YQMUAHGAEMTDSGPZNMAB4SJUYL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL/action/storage_attestation","attest_author":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL/action/author_attestation","sign_citation":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL/action/citation_signature","submit_replication":"https://pith.science/pith/YQMUAHGAEMTDSGPZNMAB4SJUYL/action/replication_record"}},"created_at":"2026-07-05T11:14:32.340269+00:00","updated_at":"2026-07-05T11:14:32.340269+00:00"}