{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FZIS5C7SSQGNSRPQRNPIPZTHXX","short_pith_number":"pith:FZIS5C7S","schema_version":"1.0","canonical_sha256":"2e512e8bf2940cd945f08b5e87e667bdd1497fc6fc56f92ec2bdd84acefbc6f6","source":{"kind":"arxiv","id":"2604.09552","version":1},"attestation_state":"computed","paper":{"title":"MCERF: Advancing Multimodal LLM Evaluation of Engineering Documentation with Enhanced Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A multimodal retrieval framework improves accuracy on engineering document questions by 41 percent relative to standard RAG.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.IR","authors_text":"Amir Mohammad Vahedi, Anna C. Doris, Daniele Grandi, Faez Ahmed, Hoang Anh Nguyen, Hongyi Xu, Kiarash Naghavi Khanghah","submitted_at":"2026-01-31T03:09:47Z","abstract_excerpt":"Engineering rulebooks and technical standards contain multimodal information like dense text, tables, and illustrations that are challenging for retrieval augmented generation (RAG) systems. Building upon the DesignQA framework [1], which relied on full-text ingestion and text-based retrieval, this work establishes a Multimodal ColPali Enhanced Retrieval and Reasoning Framework (MCERF), a system that couples a multimodal retriever with large language model reasoning for accurate and efficient question answering from engineering documents. The system employs the ColPali, which retrieves both te"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2604.09552","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-01-31T03:09:47Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"02ad5f6e574ba78d567f088960071492a2cfbb8eb91441fff54233dd80f63c10","abstract_canon_sha256":"5522691e294366b0398d3a66d0394dc7f76e5abeb560f03b1dbbd45de75b51f2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:39.899707Z","signature_b64":"b39qcKz8iHuISygPl4Fjwjo234OGu07OOVJbolWnsFVNw2l5CF/6x+705IFIwRo31OuHjRKp7uqVmbZJu3+ACQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2e512e8bf2940cd945f08b5e87e667bdd1497fc6fc56f92ec2bdd84acefbc6f6","last_reissued_at":"2026-05-28T01:04:39.899281Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:39.899281Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MCERF: Advancing Multimodal LLM Evaluation of Engineering Documentation with Enhanced Retrieval","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A multimodal retrieval framework improves accuracy on engineering document questions by 41 percent relative to standard RAG.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.IR","authors_text":"Amir Mohammad Vahedi, Anna C. Doris, Daniele Grandi, Faez Ahmed, Hoang Anh Nguyen, Hongyi Xu, Kiarash Naghavi Khanghah","submitted_at":"2026-01-31T03:09:47Z","abstract_excerpt":"Engineering rulebooks and technical standards contain multimodal information like dense text, tables, and illustrations that are challenging for retrieval augmented generation (RAG) systems. Building upon the DesignQA framework [1], which relied on full-text ingestion and text-based retrieval, this work establishes a Multimodal ColPali Enhanced Retrieval and Reasoning Framework (MCERF), a system that couples a multimodal retriever with large language model reasoning for accurate and efficient question answering from engineering documents. The system employs the ColPali, which retrieves both te"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Evaluation on the DesignQA benchmark illustrates that this system improves average accuracy across all tasks with a relative gain of +41.1% from baseline RAG best results, which is a significant improvement in multimodal and reasoning-intensive tasks without complete rulebook ingestion.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That ColPali retrieval plus the four hand-designed reasoning modes will generalize beyond the DesignQA benchmark and that the reported accuracy lift is not driven by benchmark-specific tuning or post-hoc pipeline selection.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MCERF delivers a 41.1% relative accuracy gain on the DesignQA benchmark by combining ColPali vision-language retrieval with four specialized reasoning modes and dynamic routing.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A multimodal retrieval framework improves accuracy on engineering document questions by 41 percent relative to standard RAG.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1b6871fdd970fb7672a83ebb64f05a55ebfe652f804de6e638d8364c949e70b3"},"source":{"id":"2604.09552","kind":"arxiv","version":1},"verdict":{"id":"33360ecd-64f3-46fd-8e35-97b26872d1b9","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T09:25:47.462782Z","strongest_claim":"Evaluation on the DesignQA benchmark illustrates that this system improves average accuracy across all tasks with a relative gain of +41.1% from baseline RAG best results, which is a significant improvement in multimodal and reasoning-intensive tasks without complete rulebook ingestion.","one_line_summary":"MCERF delivers a 41.1% relative accuracy gain on the DesignQA benchmark by combining ColPali vision-language retrieval with four specialized reasoning modes and dynamic routing.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That ColPali retrieval plus the four hand-designed reasoning modes will generalize beyond the DesignQA benchmark and that the reported accuracy lift is not driven by benchmark-specific tuning or post-hoc pipeline selection.","pith_extraction_headline":"A multimodal retrieval framework improves accuracy on engineering document questions by 41 percent relative to standard RAG."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.09552/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":56,"sample":[{"doi":"","year":2025,"title":"Designqa: A multimodal benchmark for evaluating large language models’ understanding of engineering documentation,","work_id":"7d79c523-5f88-43eb-b733-74b893f6c60d","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Generative Models for Multimodal Docu- ment Understanding,","work_id":"4e50a1d9-78d3-4f99-afe9-c2e6beceb903","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Layout-Aware Pre-training for Visually Rich Document Understanding,","work_id":"82e45371-d975-4d65-85e2-2b65e7ea454e","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"ColPali: Efficient Document Retrieval with Vision Language Models","work_id":"d2468d08-90dc-4690-887a-9b10a6d3574e","ref_index":4,"cited_arxiv_id":"2407.01449","is_internal_anchor":true},{"doi":"","year":2023,"title":"A Comprehensive Review of Vision- Language Models,","work_id":"cd06fc50-a71d-4455-a3a6-5c14970045cd","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":56,"snapshot_sha256":"f89c6f1f2e71c91856c9620be61ed44e0bfe11ba2d487e0d52144aeeb1eb5987","internal_anchors":4},"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":"2604.09552","created_at":"2026-05-28T01:04:39.899345+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.09552v1","created_at":"2026-05-28T01:04:39.899345+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.09552","created_at":"2026-05-28T01:04:39.899345+00:00"},{"alias_kind":"pith_short_12","alias_value":"FZIS5C7SSQGN","created_at":"2026-05-28T01:04:39.899345+00:00"},{"alias_kind":"pith_short_16","alias_value":"FZIS5C7SSQGNSRPQ","created_at":"2026-05-28T01:04:39.899345+00:00"},{"alias_kind":"pith_short_8","alias_value":"FZIS5C7S","created_at":"2026-05-28T01:04:39.899345+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX","json":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX.json","graph_json":"https://pith.science/api/pith-number/FZIS5C7SSQGNSRPQRNPIPZTHXX/graph.json","events_json":"https://pith.science/api/pith-number/FZIS5C7SSQGNSRPQRNPIPZTHXX/events.json","paper":"https://pith.science/paper/FZIS5C7S"},"agent_actions":{"view_html":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX","download_json":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX.json","view_paper":"https://pith.science/paper/FZIS5C7S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.09552&json=true","fetch_graph":"https://pith.science/api/pith-number/FZIS5C7SSQGNSRPQRNPIPZTHXX/graph.json","fetch_events":"https://pith.science/api/pith-number/FZIS5C7SSQGNSRPQRNPIPZTHXX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX/action/storage_attestation","attest_author":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX/action/author_attestation","sign_citation":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX/action/citation_signature","submit_replication":"https://pith.science/pith/FZIS5C7SSQGNSRPQRNPIPZTHXX/action/replication_record"}},"created_at":"2026-05-28T01:04:39.899345+00:00","updated_at":"2026-05-28T01:04:39.899345+00:00"}