{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:JL4XWURNUB42PWGY5WJAH2OQ2V","short_pith_number":"pith:JL4XWURN","canonical_record":{"source":{"id":"2605.14062","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-13T19:35:49Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c3645564c7242c4f82677c96bd8c8ea47728c8f497e1b34bf103d0cfac28f41a","abstract_canon_sha256":"644e02ec984469ffb8ffae3f07f3e59e05867e7125d416951616ae3cab6fdeb7"},"schema_version":"1.0"},"canonical_sha256":"4af97b522da079a7d8d8ed9203e9d0d5698345d7ac6948a05d4eed8938accd45","source":{"kind":"arxiv","id":"2605.14062","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14062","created_at":"2026-05-17T23:39:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14062v1","created_at":"2026-05-17T23:39:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14062","created_at":"2026-05-17T23:39:12Z"},{"alias_kind":"pith_short_12","alias_value":"JL4XWURNUB42","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"JL4XWURNUB42PWGY","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"JL4XWURN","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:JL4XWURNUB42PWGY5WJAH2OQ2V","target":"record","payload":{"canonical_record":{"source":{"id":"2605.14062","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-13T19:35:49Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c3645564c7242c4f82677c96bd8c8ea47728c8f497e1b34bf103d0cfac28f41a","abstract_canon_sha256":"644e02ec984469ffb8ffae3f07f3e59e05867e7125d416951616ae3cab6fdeb7"},"schema_version":"1.0"},"canonical_sha256":"4af97b522da079a7d8d8ed9203e9d0d5698345d7ac6948a05d4eed8938accd45","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:12.516443Z","signature_b64":"kM06Ay+xr/i6jY5uZv1PgkxQ6CeabSTydbBhgCtZGPFCEHXb8Q7SuK4V+jCKrNsNfRBQOgQluv3g2rX7KhxGCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4af97b522da079a7d8d8ed9203e9d0d5698345d7ac6948a05d4eed8938accd45","last_reissued_at":"2026-05-17T23:39:12.515915Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:12.515915Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.14062","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:12Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"NRZ367c8iXxGM4i8y347SXpsG8AQfhT9yChRT5BfDs6ehBx4rKOiRJndu7WlTJbO2Z++6degBQgiY7E2m47hCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T14:00:00.306695Z"},"content_sha256":"2cd2b98c5bb0c6008530364f7ab500a71a13837da6f6c077de744d1c9f26cd0b","schema_version":"1.0","event_id":"sha256:2cd2b98c5bb0c6008530364f7ab500a71a13837da6f6c077de744d1c9f26cd0b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:JL4XWURNUB42PWGY5WJAH2OQ2V","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Know When To Fold 'Em: Token-Efficient LLM Synthetic Data Generation via Multi-Stage In-Flight Rejection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MSIFR cuts token use in LLM synthetic data generation by 11-77% through early rejection of faulty outputs at intermediate stages.","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Anjir Ahmed Chowdhury, Feng Yan, Syed Zawad","submitted_at":"2026-05-13T19:35:49Z","abstract_excerpt":"While synthetic data generation with large language models (LLMs) is widely used in post-training pipelines, existing approaches typically generate full outputs before applying quality filters, leading to substantial token waste on samples that are ultimately discarded. To address this, we propose Multi-Stage In-Flight Rejection (MSIFR), a lightweight, training-free framework that detects and terminates low-quality generation trajectories at intermediate checkpoints before they reach full completion. MSIFR decomposes the generation process into sequential stages and applies fast rule-based val"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MSIFR reduces token consumption by 11%-77% as a standalone method, and up to 78.2% when combined with early-exit methods, while preserving or improving evaluation accuracy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The fast rule-based validators at intermediate stages can reliably identify low-quality trajectories without discarding many high-quality samples or missing many low-quality ones that would waste tokens if completed.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MSIFR stops faulty LLM generations early via staged rule-based checks, reducing token consumption 11-78% with no accuracy loss.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MSIFR cuts token use in LLM synthetic data generation by 11-77% through early rejection of faulty outputs at intermediate stages.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"804fe164780d89b7964df4466709ef25cf610b18e1ca6c07ddf8bae0ec47d636"},"source":{"id":"2605.14062","kind":"arxiv","version":1},"verdict":{"id":"05ba699e-d385-4dbf-bfd7-945ec0c9975e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T05:00:38.530816Z","strongest_claim":"MSIFR reduces token consumption by 11%-77% as a standalone method, and up to 78.2% when combined with early-exit methods, while preserving or improving evaluation accuracy.","one_line_summary":"MSIFR stops faulty LLM generations early via staged rule-based checks, reducing token consumption 11-78% with no accuracy loss.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The fast rule-based validators at intermediate stages can reliably identify low-quality trajectories without discarding many high-quality samples or missing many low-quality ones that would waste tokens if completed.","pith_extraction_headline":"MSIFR cuts token use in LLM synthetic data generation by 11-77% through early rejection of faulty outputs at intermediate stages."},"references":{"count":49,"sample":[{"doi":"","year":null,"title":"arXiv preprint arXiv:2501.12345 , year=","work_id":"216add13-ee85-43c4-845b-e6c521435fd8","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles , year=","work_id":"9a3c3fdb-ec54-4e64-9779-63c57fad506c","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1991,"title":"Probability with Martingales , author=. 1991 , publisher=","work_id":"4280221a-48ba-4531-8102-193775c057d9","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Accelerating","work_id":"f1329995-d6b5-4cab-bca7-e50a0b3254f9","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2005,"title":"Algorithmic Learning in a Random World , author=. 2005 , publisher=","work_id":"6b532a4c-7e17-4325-82ab-596b46df161d","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":49,"snapshot_sha256":"98ca5898192c581661f27a27b648797bd0165bc2830bf3350ac899b03b76e1b8","internal_anchors":17},"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"},"verdict_id":"05ba699e-d385-4dbf-bfd7-945ec0c9975e"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:12Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"cHGURYchZGtBp8K098TqjTd49I9O7WggPRuuKYk6Lml7bCXI4KZgfndyOyBCq3+C24njxmhI3zkWNCcKXDMPAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T14:00:00.307185Z"},"content_sha256":"15b919b9eb7c40135caf89f23ee4d35e96f0d0c217db6a75b1f22246a5a0e584","schema_version":"1.0","event_id":"sha256:15b919b9eb7c40135caf89f23ee4d35e96f0d0c217db6a75b1f22246a5a0e584"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JL4XWURNUB42PWGY5WJAH2OQ2V/bundle.json","state_url":"https://pith.science/pith/JL4XWURNUB42PWGY5WJAH2OQ2V/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JL4XWURNUB42PWGY5WJAH2OQ2V/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-01T14:00:00Z","links":{"resolver":"https://pith.science/pith/JL4XWURNUB42PWGY5WJAH2OQ2V","bundle":"https://pith.science/pith/JL4XWURNUB42PWGY5WJAH2OQ2V/bundle.json","state":"https://pith.science/pith/JL4XWURNUB42PWGY5WJAH2OQ2V/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JL4XWURNUB42PWGY5WJAH2OQ2V/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:JL4XWURNUB42PWGY5WJAH2OQ2V","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":"644e02ec984469ffb8ffae3f07f3e59e05867e7125d416951616ae3cab6fdeb7","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-13T19:35:49Z","title_canon_sha256":"c3645564c7242c4f82677c96bd8c8ea47728c8f497e1b34bf103d0cfac28f41a"},"schema_version":"1.0","source":{"id":"2605.14062","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14062","created_at":"2026-05-17T23:39:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14062v1","created_at":"2026-05-17T23:39:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14062","created_at":"2026-05-17T23:39:12Z"},{"alias_kind":"pith_short_12","alias_value":"JL4XWURNUB42","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"JL4XWURNUB42PWGY","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"JL4XWURN","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:15b919b9eb7c40135caf89f23ee4d35e96f0d0c217db6a75b1f22246a5a0e584","target":"graph","created_at":"2026-05-17T23:39:12Z","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":"MSIFR reduces token consumption by 11%-77% as a standalone method, and up to 78.2% when combined with early-exit methods, while preserving or improving evaluation accuracy."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The fast rule-based validators at intermediate stages can reliably identify low-quality trajectories without discarding many high-quality samples or missing many low-quality ones that would waste tokens if completed."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"MSIFR stops faulty LLM generations early via staged rule-based checks, reducing token consumption 11-78% with no accuracy loss."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"MSIFR cuts token use in LLM synthetic data generation by 11-77% through early rejection of faulty outputs at intermediate stages."}],"snapshot_sha256":"804fe164780d89b7964df4466709ef25cf610b18e1ca6c07ddf8bae0ec47d636"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"While synthetic data generation with large language models (LLMs) is widely used in post-training pipelines, existing approaches typically generate full outputs before applying quality filters, leading to substantial token waste on samples that are ultimately discarded. To address this, we propose Multi-Stage In-Flight Rejection (MSIFR), a lightweight, training-free framework that detects and terminates low-quality generation trajectories at intermediate checkpoints before they reach full completion. MSIFR decomposes the generation process into sequential stages and applies fast rule-based val","authors_text":"Anjir Ahmed Chowdhury, Feng Yan, Syed Zawad","cross_cats":["cs.CL"],"headline":"MSIFR cuts token use in LLM synthetic data generation by 11-77% through early rejection of faulty outputs at intermediate stages.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-13T19:35:49Z","title":"Know When To Fold 'Em: Token-Efficient LLM Synthetic Data Generation via Multi-Stage In-Flight Rejection"},"references":{"count":49,"internal_anchors":17,"resolved_work":49,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"arXiv preprint arXiv:2501.12345 , year=","work_id":"216add13-ee85-43c4-845b-e6c521435fd8","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles , year=","work_id":"9a3c3fdb-ec54-4e64-9779-63c57fad506c","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Probability with Martingales , author=. 1991 , publisher=","work_id":"4280221a-48ba-4531-8102-193775c057d9","year":1991},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Accelerating","work_id":"f1329995-d6b5-4cab-bca7-e50a0b3254f9","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Algorithmic Learning in a Random World , author=. 2005 , publisher=","work_id":"6b532a4c-7e17-4325-82ab-596b46df161d","year":2005}],"snapshot_sha256":"98ca5898192c581661f27a27b648797bd0165bc2830bf3350ac899b03b76e1b8"},"source":{"id":"2605.14062","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T05:00:38.530816Z","id":"05ba699e-d385-4dbf-bfd7-945ec0c9975e","model_set":{"reader":"grok-4.3"},"one_line_summary":"MSIFR stops faulty LLM generations early via staged rule-based checks, reducing token consumption 11-78% with no accuracy loss.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"MSIFR cuts token use in LLM synthetic data generation by 11-77% through early rejection of faulty outputs at intermediate stages.","strongest_claim":"MSIFR reduces token consumption by 11%-77% as a standalone method, and up to 78.2% when combined with early-exit methods, while preserving or improving evaluation accuracy.","weakest_assumption":"The fast rule-based validators at intermediate stages can reliably identify low-quality trajectories without discarding many high-quality samples or missing many low-quality ones that would waste tokens if completed."}},"verdict_id":"05ba699e-d385-4dbf-bfd7-945ec0c9975e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:2cd2b98c5bb0c6008530364f7ab500a71a13837da6f6c077de744d1c9f26cd0b","target":"record","created_at":"2026-05-17T23:39:12Z","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":"644e02ec984469ffb8ffae3f07f3e59e05867e7125d416951616ae3cab6fdeb7","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-13T19:35:49Z","title_canon_sha256":"c3645564c7242c4f82677c96bd8c8ea47728c8f497e1b34bf103d0cfac28f41a"},"schema_version":"1.0","source":{"id":"2605.14062","kind":"arxiv","version":1}},"canonical_sha256":"4af97b522da079a7d8d8ed9203e9d0d5698345d7ac6948a05d4eed8938accd45","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4af97b522da079a7d8d8ed9203e9d0d5698345d7ac6948a05d4eed8938accd45","first_computed_at":"2026-05-17T23:39:12.515915Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:12.515915Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"kM06Ay+xr/i6jY5uZv1PgkxQ6CeabSTydbBhgCtZGPFCEHXb8Q7SuK4V+jCKrNsNfRBQOgQluv3g2rX7KhxGCg==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:12.516443Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14062","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2cd2b98c5bb0c6008530364f7ab500a71a13837da6f6c077de744d1c9f26cd0b","sha256:15b919b9eb7c40135caf89f23ee4d35e96f0d0c217db6a75b1f22246a5a0e584"],"state_sha256":"726a4fa7e03472d89b47b55f5e7f368cd057af365899d3d6e4feb0edff5ff978"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jobIt3+jjdT+V8QloUX/XT2d4jB019gZNhjnxFp8jkcnRBZMVciOvSXO06+GGt/CiUXHK1zY9TB/e2yYHN8XCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-01T14:00:00.309463Z","bundle_sha256":"404737046d9503e7925875605b3cf0c2dd13bf935c880c7ceb072dff1a546bd2"}}