{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:77ABU4DH6AOOQG3MZ2K5VSA7MQ","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":"592a532a7cc2094cb4cf0ef1e6f7f2da87901236124d12dad3c23475b383bb3d","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SE","submitted_at":"2026-05-17T17:38:17Z","title_canon_sha256":"5605cc685a86d0ae2548fada9dbc1ca5b84d98ffe19dffd9dfa6b056758910b6"},"schema_version":"1.0","source":{"id":"2605.17558","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17558","created_at":"2026-05-20T00:04:45Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17558v1","created_at":"2026-05-20T00:04:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17558","created_at":"2026-05-20T00:04:45Z"},{"alias_kind":"pith_short_12","alias_value":"77ABU4DH6AOO","created_at":"2026-05-20T00:04:45Z"},{"alias_kind":"pith_short_16","alias_value":"77ABU4DH6AOOQG3M","created_at":"2026-05-20T00:04:45Z"},{"alias_kind":"pith_short_8","alias_value":"77ABU4DH","created_at":"2026-05-20T00:04:45Z"}],"graph_snapshots":[{"event_id":"sha256:882ef0465ee1f66137e3302570153220d1bb611f1f4a720efb95e51d7144b896","target":"graph","created_at":"2026-05-20T00:04:45Z","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":"A 4B-parameter model trained with GRPO on FireFly matches Claude Sonnet 4.6 on our held-out test set and shows improvements on multiple tool-calling benchmarks including Tau2-Bench, MCPMark, and MCP-Atlas."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that backward synthesis from observed API outcomes produces tasks whose labels remain correct and useful when the same tasks are later executed in the retrieval-augmented simulator or on live APIs (abstract, paragraph describing the pipeline inversion)."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"FireFly inverts task synthesis by exploring real MCP servers first via pairwise tool graphs and sub-DAG sampling, then generates 5,144 verified tasks backward from outcomes to train a 4B model that matches Claude Sonnet 4.6 on tool-calling benchmarks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"FireFly inverts the data synthesis pipeline to generate verified tool-calling trajectories directly from real API explorations."}],"snapshot_sha256":"bf71460c6987eb65a7ece163769354812f38b09e2353f8e536d9f4e8359cb66b"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"a31ef12599958e22a50234bf187b0f56c8f45dbff71ab62f7f55ed4aa5a9bdaf"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.574986Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T22:31:06.514378Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.603011Z","status":"skipped","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.536094Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.17558/integrity.json","findings":[],"snapshot_sha256":"23031d6f7fc0b70c0c3e9f8593f06c944712f6a2a270d10084bc53d109f0d44d","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training tool-calling agents requires large-scale trajectory data with verifiable labels, yet existing approaches either synthesize environments that diverge from real API behavior or generate tasks without ground-truth outcomes for verification. We present FireFly, a pipeline for generating verified tool-call data from real-world MCP servers. Our key insight is to invert the standard synthesis pipeline: rather than generating tasks and hoping they are solvable, we first let a strong LLM explore real APIs along graph-guided DAG structures, then synthesize tasks backward from observed outcomes,","authors_text":"Benoit Dumoulin, Chen Luo, Dakuo Wang, Hanqing Lu, Hui Liu, Jing Huang, Jiri Gesi, Qi He, Xianfeng Tang, Yimeng Zhang, Yingzhou Lu, Yisi Sang, Yuxuan Lu, Zhenwei Dai, Ziyi Wang","cross_cats":["cs.CL"],"headline":"FireFly inverts the data synthesis pipeline to generate verified tool-calling trajectories directly from real API explorations.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SE","submitted_at":"2026-05-17T17:38:17Z","title":"Firefly: Illuminating Large-Scale Verified Tool-Call Data Generation from Real APIs"},"references":{"count":25,"internal_anchors":10,"resolved_work":25,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers, May 2026","work_id":"f6630c64-4616-44e2-98ca-63891932848f","year":2026},{"cited_arxiv_id":"2506.07982","doi":"","is_internal_anchor":true,"ref_index":2,"title":"$\\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment","work_id":"3a498b1a-455f-4667-b572-c5216c99a89c","year":2025},{"cited_arxiv_id":"2107.03374","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Scaling Agent Learning via Experience Synthesis, November 2025","work_id":"c4ed3596-4df7-4a87-bc6b-660810a70685","year":2025},{"cited_arxiv_id":"2304.08244","doi":"","is_internal_anchor":true,"ref_index":5,"title":"API-Bank: A Comprehensive Benchmark for Tool-Augmented LLMs","work_id":"a20d9332-ab34-485c-a060-1ba47cc98930","year":2023}],"snapshot_sha256":"19a5624ae0a7d4c4983df0b8fe4a12fe226931faf83b0dd0065460f2ebda87d4"},"source":{"id":"2605.17558","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T22:26:02.189614Z","id":"8e72320f-fb29-47a2-8002-b3a38e738155","model_set":{"reader":"grok-4.3"},"one_line_summary":"FireFly inverts task synthesis by exploring real MCP servers first via pairwise tool graphs and sub-DAG sampling, then generates 5,144 verified tasks backward from outcomes to train a 4B model that matches Claude Sonnet 4.6 on tool-calling benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"FireFly inverts the data synthesis pipeline to generate verified tool-calling trajectories directly from real API explorations.","strongest_claim":"A 4B-parameter model trained with GRPO on FireFly matches Claude Sonnet 4.6 on our held-out test set and shows improvements on multiple tool-calling benchmarks including Tau2-Bench, MCPMark, and MCP-Atlas.","weakest_assumption":"The assumption that backward synthesis from observed API outcomes produces tasks whose labels remain correct and useful when the same tasks are later executed in the retrieval-augmented simulator or on live APIs (abstract, paragraph describing the pipeline inversion)."}},"verdict_id":"8e72320f-fb29-47a2-8002-b3a38e738155"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b48bf5ea2cb24b25620b09e5940ccee4740eb3b9b8cfa2c63cdc119b2110b1d0","target":"record","created_at":"2026-05-20T00:04:45Z","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":"592a532a7cc2094cb4cf0ef1e6f7f2da87901236124d12dad3c23475b383bb3d","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SE","submitted_at":"2026-05-17T17:38:17Z","title_canon_sha256":"5605cc685a86d0ae2548fada9dbc1ca5b84d98ffe19dffd9dfa6b056758910b6"},"schema_version":"1.0","source":{"id":"2605.17558","kind":"arxiv","version":1}},"canonical_sha256":"ffc01a7067f01ce81b6cce95dac81f642fb04a3ef22b72379075eb9836c1f7d6","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ffc01a7067f01ce81b6cce95dac81f642fb04a3ef22b72379075eb9836c1f7d6","first_computed_at":"2026-05-20T00:04:45.792098Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:45.792098Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"J0UsGss/RHqiv41El7JCDtNN08IgapIXAJBgUYv1aANVyqqHt7tF6OKU8O8npr30QGxqY1nM7oE+kiehCn7TCg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:45.793048Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.17558","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b48bf5ea2cb24b25620b09e5940ccee4740eb3b9b8cfa2c63cdc119b2110b1d0","sha256:882ef0465ee1f66137e3302570153220d1bb611f1f4a720efb95e51d7144b896"],"state_sha256":"83336d8a92a6612ce1b04009eeb49d9b0da138b24aec7f4ec30f65e257a93771"}