{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:O7ERIZCRAMKRZS52XQH47STJPG","short_pith_number":"pith:O7ERIZCR","canonical_record":{"source":{"id":"2604.17943","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-20T08:22:15Z","cross_cats_sorted":[],"title_canon_sha256":"cc2402e07d10e51b85970b9a2ef303ed2aa16bc7f1569f7df2695ecefb0f9328","abstract_canon_sha256":"295df0d29c5b8a9d1ede765d6cf4940032fcb86256c9adb34eebd7db67dd1387"},"schema_version":"1.0"},"canonical_sha256":"77c914645103151ccbbabc0fcfca69799aae45aea6d8d999217dffc47cbd54d9","source":{"kind":"arxiv","id":"2604.17943","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.17943","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"arxiv_version","alias_value":"2604.17943v2","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.17943","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"pith_short_12","alias_value":"O7ERIZCRAMKR","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"pith_short_16","alias_value":"O7ERIZCRAMKRZS52","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"pith_short_8","alias_value":"O7ERIZCR","created_at":"2026-05-28T01:04:40Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:O7ERIZCRAMKRZS52XQH47STJPG","target":"record","payload":{"canonical_record":{"source":{"id":"2604.17943","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-20T08:22:15Z","cross_cats_sorted":[],"title_canon_sha256":"cc2402e07d10e51b85970b9a2ef303ed2aa16bc7f1569f7df2695ecefb0f9328","abstract_canon_sha256":"295df0d29c5b8a9d1ede765d6cf4940032fcb86256c9adb34eebd7db67dd1387"},"schema_version":"1.0"},"canonical_sha256":"77c914645103151ccbbabc0fcfca69799aae45aea6d8d999217dffc47cbd54d9","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:40.374578Z","signature_b64":"HZc1NzFix9ZmQrL423Dxw7HWMLLt8wOoajyyXj68wp06RGlRcE8mbZLAeeiT6JunCMlq4+r7uI0yvDQTfIibCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"77c914645103151ccbbabc0fcfca69799aae45aea6d8d999217dffc47cbd54d9","last_reissued_at":"2026-05-28T01:04:40.374120Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:40.374120Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.17943","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-28T01:04:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9Jqf1bbHmVxy+19NgnyNpVHFT3ecSZKWh+NF3cN14twdeqNv53fEFcPPdnDPYVqQ0qVcGdvC6sEhAEvKN0tlDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-06T23:34:12.486866Z"},"content_sha256":"74bc4950c4f84f86b82545fa4acaa624b41a097fc2f4a588336dcfb57a5664df","schema_version":"1.0","event_id":"sha256:74bc4950c4f84f86b82545fa4acaa624b41a097fc2f4a588336dcfb57a5664df"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:O7ERIZCRAMKRZS52XQH47STJPG","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"A Benchmark Construction and Evaluation Framework for Specialist Domains: Case Study on Defense-related Documents","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"A model fine-tuned on the DoRA benchmark achieves up to 26% higher QA success and 47% lower hallucination rates on defense documents.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aarya Bodhankar, Aditya Joshi, Bao Gia Doan, Flora Salim, Oscar Leslie, Pantelis Elinas, Tom Marchant","submitted_at":"2026-04-20T08:22:15Z","abstract_excerpt":"RAG-based question-answering (QA) in specialist domains faces a cold-start problem: lack of evaluative benchmarks and absence of labeled data for post-training. We present DoRA (Domain-oriented RAG Assessment), a novel benchmark construction and evaluation framework using only a small set of specialist domain documents. DoRA systematically generates synthetic QA training and evaluation datasets with auditable evidence across five domain-specific intents. To mitigate same-pipeline circularity, DoRA's training and test splits use different LLM families (Claude Sonnet for training; GPT-4o for tes"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"a model trained on DoRA (DoRA SFT) yields large gains over the base model (Llama3.1-8B-Instruct): up to 26% improvement in QA task success, while reducing the hallucination rate by 47% in RAG faithfulness scores","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The synthetic, intent-conditioned QA pairs and curated evidence passages faithfully represent the distribution and attribution challenges of real user queries on defense documents without introducing generation artifacts or selection bias.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DoRA is a new synthetic benchmark for RAG-based QA on defense documents where fine-tuning Llama3.1-8B-Instruct on it improves task success by up to 26% and cuts hallucination rates by 47%.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A model fine-tuned on the DoRA benchmark achieves up to 26% higher QA success and 47% lower hallucination rates on defense documents.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"482036ac4700af60f5c832a1b795d8c2ada3645f5f62d33f49e1f44ad1fadfe7"},"source":{"id":"2604.17943","kind":"arxiv","version":2},"verdict":{"id":"b5b43763-29be-405c-ba10-e2be079ca664","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T04:47:26.021886Z","strongest_claim":"a model trained on DoRA (DoRA SFT) yields large gains over the base model (Llama3.1-8B-Instruct): up to 26% improvement in QA task success, while reducing the hallucination rate by 47% in RAG faithfulness scores","one_line_summary":"DoRA is a new synthetic benchmark for RAG-based QA on defense documents where fine-tuning Llama3.1-8B-Instruct on it improves task success by up to 26% and cuts hallucination rates by 47%.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The synthetic, intent-conditioned QA pairs and curated evidence passages faithfully represent the distribution and attribution challenges of real user queries on defense documents without introducing generation artifacts or selection bias.","pith_extraction_headline":"A model fine-tuned on the DoRA benchmark achieves up to 26% higher QA success and 47% lower hallucination rates on defense documents."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.17943/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-20T04:35:44.875670Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"ff7a7d005b574324282a92ec99238b8f1638a3c3bc1cc4888955df8311be2c71"},"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"},"verdict_id":"b5b43763-29be-405c-ba10-e2be079ca664"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-28T01:04:40Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XK/waHv/rmbfXFrQQc4GvfG0NX+uOcgK3P8Neg1iZH2HDkuENkNNZfUn523erlul94ZEXi2UopFKC/jX6SM9BA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-06T23:34:12.487702Z"},"content_sha256":"bf5cf6630dd94d28b07afcd29c6cdcd409bec87c77422602b2f0be6db5e08ecd","schema_version":"1.0","event_id":"sha256:bf5cf6630dd94d28b07afcd29c6cdcd409bec87c77422602b2f0be6db5e08ecd"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/O7ERIZCRAMKRZS52XQH47STJPG/bundle.json","state_url":"https://pith.science/pith/O7ERIZCRAMKRZS52XQH47STJPG/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/O7ERIZCRAMKRZS52XQH47STJPG/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":"2026-06-06T23:34:12Z","links":{"resolver":"https://pith.science/pith/O7ERIZCRAMKRZS52XQH47STJPG","bundle":"https://pith.science/pith/O7ERIZCRAMKRZS52XQH47STJPG/bundle.json","state":"https://pith.science/pith/O7ERIZCRAMKRZS52XQH47STJPG/state.json","well_known_bundle":"https://pith.science/.well-known/pith/O7ERIZCRAMKRZS52XQH47STJPG/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:O7ERIZCRAMKRZS52XQH47STJPG","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":"295df0d29c5b8a9d1ede765d6cf4940032fcb86256c9adb34eebd7db67dd1387","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-20T08:22:15Z","title_canon_sha256":"cc2402e07d10e51b85970b9a2ef303ed2aa16bc7f1569f7df2695ecefb0f9328"},"schema_version":"1.0","source":{"id":"2604.17943","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.17943","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"arxiv_version","alias_value":"2604.17943v2","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.17943","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"pith_short_12","alias_value":"O7ERIZCRAMKR","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"pith_short_16","alias_value":"O7ERIZCRAMKRZS52","created_at":"2026-05-28T01:04:40Z"},{"alias_kind":"pith_short_8","alias_value":"O7ERIZCR","created_at":"2026-05-28T01:04:40Z"}],"graph_snapshots":[{"event_id":"sha256:bf5cf6630dd94d28b07afcd29c6cdcd409bec87c77422602b2f0be6db5e08ecd","target":"graph","created_at":"2026-05-28T01:04:40Z","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 model trained on DoRA (DoRA SFT) yields large gains over the base model (Llama3.1-8B-Instruct): up to 26% improvement in QA task success, while reducing the hallucination rate by 47% in RAG faithfulness scores"},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The synthetic, intent-conditioned QA pairs and curated evidence passages faithfully represent the distribution and attribution challenges of real user queries on defense documents without introducing generation artifacts or selection bias."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DoRA is a new synthetic benchmark for RAG-based QA on defense documents where fine-tuning Llama3.1-8B-Instruct on it improves task success by up to 26% and cuts hallucination rates by 47%."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A model fine-tuned on the DoRA benchmark achieves up to 26% higher QA success and 47% lower hallucination rates on defense documents."}],"snapshot_sha256":"482036ac4700af60f5c832a1b795d8c2ada3645f5f62d33f49e1f44ad1fadfe7"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-20T04:35:44.875670Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.17943/integrity.json","findings":[],"snapshot_sha256":"ff7a7d005b574324282a92ec99238b8f1638a3c3bc1cc4888955df8311be2c71","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"RAG-based question-answering (QA) in specialist domains faces a cold-start problem: lack of evaluative benchmarks and absence of labeled data for post-training. We present DoRA (Domain-oriented RAG Assessment), a novel benchmark construction and evaluation framework using only a small set of specialist domain documents. DoRA systematically generates synthetic QA training and evaluation datasets with auditable evidence across five domain-specific intents. To mitigate same-pipeline circularity, DoRA's training and test splits use different LLM families (Claude Sonnet for training; GPT-4o for tes","authors_text":"Aarya Bodhankar, Aditya Joshi, Bao Gia Doan, Flora Salim, Oscar Leslie, Pantelis Elinas, Tom Marchant","cross_cats":[],"headline":"A model fine-tuned on the DoRA benchmark achieves up to 26% higher QA success and 47% lower hallucination rates on defense documents.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-20T08:22:15Z","title":"A Benchmark Construction and Evaluation Framework for Specialist Domains: Case Study on Defense-related Documents"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.17943","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T04:47:26.021886Z","id":"b5b43763-29be-405c-ba10-e2be079ca664","model_set":{"reader":"grok-4.3"},"one_line_summary":"DoRA is a new synthetic benchmark for RAG-based QA on defense documents where fine-tuning Llama3.1-8B-Instruct on it improves task success by up to 26% and cuts hallucination rates by 47%.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A model fine-tuned on the DoRA benchmark achieves up to 26% higher QA success and 47% lower hallucination rates on defense documents.","strongest_claim":"a model trained on DoRA (DoRA SFT) yields large gains over the base model (Llama3.1-8B-Instruct): up to 26% improvement in QA task success, while reducing the hallucination rate by 47% in RAG faithfulness scores","weakest_assumption":"The synthetic, intent-conditioned QA pairs and curated evidence passages faithfully represent the distribution and attribution challenges of real user queries on defense documents without introducing generation artifacts or selection bias."}},"verdict_id":"b5b43763-29be-405c-ba10-e2be079ca664"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:74bc4950c4f84f86b82545fa4acaa624b41a097fc2f4a588336dcfb57a5664df","target":"record","created_at":"2026-05-28T01:04:40Z","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":"295df0d29c5b8a9d1ede765d6cf4940032fcb86256c9adb34eebd7db67dd1387","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-20T08:22:15Z","title_canon_sha256":"cc2402e07d10e51b85970b9a2ef303ed2aa16bc7f1569f7df2695ecefb0f9328"},"schema_version":"1.0","source":{"id":"2604.17943","kind":"arxiv","version":2}},"canonical_sha256":"77c914645103151ccbbabc0fcfca69799aae45aea6d8d999217dffc47cbd54d9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"77c914645103151ccbbabc0fcfca69799aae45aea6d8d999217dffc47cbd54d9","first_computed_at":"2026-05-28T01:04:40.374120Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:04:40.374120Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"HZc1NzFix9ZmQrL423Dxw7HWMLLt8wOoajyyXj68wp06RGlRcE8mbZLAeeiT6JunCMlq4+r7uI0yvDQTfIibCw==","signature_status":"signed_v1","signed_at":"2026-05-28T01:04:40.374578Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.17943","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:74bc4950c4f84f86b82545fa4acaa624b41a097fc2f4a588336dcfb57a5664df","sha256:bf5cf6630dd94d28b07afcd29c6cdcd409bec87c77422602b2f0be6db5e08ecd"],"state_sha256":"2409407bdfdb5f15514a44f4410c1e8c9258321a18b0e306960fea5b6f8665f5"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HTPWSJORMxa2QyAuKvZ5BryWMiCu3k6jua1AvwWke3FFOSughnDEeBvtD2sKVeOyCyf8YkVWrkW1GdETfHtFCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-06T23:34:12.492046Z","bundle_sha256":"c12456240eaa5cc9ccac5be28d0dddfced0ef11b9173abfd9a7664484838f131"}}