{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:4KLRREEZFUVP2F4MJLKVBWDBXD","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":"b7f7572a0ac91b4a8ecddc72be10fff7b46c9165c93d3dda18a63b7a7fcbe07d","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-06T05:04:00Z","title_canon_sha256":"383b2854a512b937173a50281fa1a995b501da743652bd2f9f889b3bf4894b2b"},"schema_version":"1.0","source":{"id":"2508.04086","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.04086","created_at":"2026-06-19T16:11:12Z"},{"alias_kind":"arxiv_version","alias_value":"2508.04086v3","created_at":"2026-06-19T16:11:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.04086","created_at":"2026-06-19T16:11:12Z"},{"alias_kind":"pith_short_12","alias_value":"4KLRREEZFUVP","created_at":"2026-06-19T16:11:12Z"},{"alias_kind":"pith_short_16","alias_value":"4KLRREEZFUVP2F4M","created_at":"2026-06-19T16:11:12Z"},{"alias_kind":"pith_short_8","alias_value":"4KLRREEZ","created_at":"2026-06-19T16:11:12Z"}],"graph_snapshots":[{"event_id":"sha256:3afc7bdaf4351e7f008d9cc3758b5b4b36f4a2e17bcada4ef2af4d6b3a15fe4c","target":"graph","created_at":"2026-06-19T16:11:12Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Experiments show that ToolGrad models outperform those trained on expensive baseline datasets and proprietary LLMs."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the iterative process guided by textual gradients consistently produces valid and complex tool-use chains that generalize well to real user queries."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ToolGrad inverts the standard tool-use dataset synthesis process by constructing valid tool chains first with textual gradients, producing a high-quality 500-example dataset with near-perfect validity and superior model performance."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"ToolGrad generates superior tool-use training data by building valid chains first with textual gradients before creating queries."}],"snapshot_sha256":"fd9909475f388fdb1765e9721d6ff5b5f3f52037511fef0a19cef40651da75b7"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"daae8f819e1046ec10cddd5ed7bc9e6dc291025fa7e8edb1146e9b3246d4f5c0"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2508.04086/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Prior work synthesizes tool-use LLM datasets by first generating a user query, followed by complex tool-use annotations like depth-first search (DFS). This leads to inevitable annotation failures and low efficiency in data generation. We introduce ToolGrad, an agentic framework that inverts this paradigm. ToolGrad first constructs valid tool-use chains through an iterative process guided by textual \"gradients\", and then synthesizes corresponding user queries. This \"answer-first\" approach led to ToolGrad-500, a dataset generated with more complex tool use, lower cost, and almost 100% pass rate.","authors_text":"Haoyu Zhang, Jingtao Zhou, Kohei Uehara, Lin Gu, Ruofei Du, Tatsuya Harada, Zheng Xu, Zhongyi Zhou","cross_cats":[],"headline":"ToolGrad generates superior tool-use training data by building valid chains first with textual gradients before creating queries.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-06T05:04:00Z","title":"ToolGrad: Efficient Tool-use Dataset Generation with Textual \"Gradients\""},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.04086","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-19T01:07:59.573950Z","id":"04a25205-0fe6-4b58-a662-fc914a67685c","model_set":{"reader":"grok-4.3"},"one_line_summary":"ToolGrad inverts the standard tool-use dataset synthesis process by constructing valid tool chains first with textual gradients, producing a high-quality 500-example dataset with near-perfect validity and superior model performance.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"ToolGrad generates superior tool-use training data by building valid chains first with textual gradients before creating queries.","strongest_claim":"Experiments show that ToolGrad models outperform those trained on expensive baseline datasets and proprietary LLMs.","weakest_assumption":"That the iterative process guided by textual gradients consistently produces valid and complex tool-use chains that generalize well to real user queries."}},"verdict_id":"04a25205-0fe6-4b58-a662-fc914a67685c"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:30afd265e5eb0f27231315ecaee024c05a93ee0c9f1049358452e99dfeff40a1","target":"record","created_at":"2026-06-19T16:11:12Z","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":"b7f7572a0ac91b4a8ecddc72be10fff7b46c9165c93d3dda18a63b7a7fcbe07d","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-06T05:04:00Z","title_canon_sha256":"383b2854a512b937173a50281fa1a995b501da743652bd2f9f889b3bf4894b2b"},"schema_version":"1.0","source":{"id":"2508.04086","kind":"arxiv","version":3}},"canonical_sha256":"e2971890992d2afd178c4ad550d861b8ed46163cd378709957b2d3f4e5434af3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e2971890992d2afd178c4ad550d861b8ed46163cd378709957b2d3f4e5434af3","first_computed_at":"2026-06-19T16:11:12.063344Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-19T16:11:12.063344Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Y9rNKoQC1sZypuF49c2V73bHbHSNaGzGWoYg8naX20G7K+SyMQqqSL1BI/T+rO2ZYO3dHZnKXwzO+rspi2zjBg==","signature_status":"signed_v1","signed_at":"2026-06-19T16:11:12.063735Z","signed_message":"canonical_sha256_bytes"},"source_id":"2508.04086","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:30afd265e5eb0f27231315ecaee024c05a93ee0c9f1049358452e99dfeff40a1","sha256:3afc7bdaf4351e7f008d9cc3758b5b4b36f4a2e17bcada4ef2af4d6b3a15fe4c"],"state_sha256":"ccb510b6fbe7a75643f1ccbf8e56adfbe5c27166afa56901d3f22ca2bb20e755"}