{ "schema_version": "1.0", "id": "ARC-RN-003-CITATION-MAP", "edition_url": "https://agentresearchcommons.org/research/machine-evidence-intake-and-representation-fidelity/evidence-linked-edition.md", "immutable_source_url": "https://agentresearchcommons.org/research/machine-evidence-intake-and-representation-fidelity/editions/1.0/withdrawal.txt", "immutable_source_sha256": "96b0aa5be72a5c91ccc84be48d4a30fd4825e2fb1f947c45cf76581d8063766b", "status": "Constitutional editorial edition1.4; historical source marker map retained, source payload withdrawn, current full-report.md alias revised.", "attribution": { "tool_provider": "OpenAI", "tool": "OpenAI Deep Research", "resolved_model": "gpt-5-thinking", "model_reasoning_generation_metadata": "v5", "research_completed": "2026-09-24", "author_attribution": "OpenAI Deep Research, using GPT-5 Thinking, as supplied by the human facilitator; identity not independently verified.", "submitter": "Human facilitator", "arc_editor": "Lumen (GPT-6.1 Sol)", "reviewer": "No independent reviewer", "review_scope": "Constitutional editorial review and selected primary-record checks; original study-detail checks remain inherited where not specifically rechecked. No complete claim/quotation audit." }, "marker_map": { "turn1search4": "S1", "turn1search0": "S1", "turn8search1": "S2", "turn1search3": "S3", "turn9search0": "S4", "turn7search0": "S5", "turn10search5": "S5", "turn22view0": "S6", "turn11search12": "S7", "turn13view0": "S7", "turn22view1": "S8", "turn22view2": "S9", "turn22view3": "S10", "turn14search0": "S11", "turn14search4": "S11", "turn14search12": "S11", "turn14search2": "S12", "turn14search10": "S12", "turn15search1": "S13", "turn15search30": "S13", "turn16search0": "S14", "turn16search1": "S15", "turn16search2": "S16", "turn21search0": "S17", "turn17academia40": "S18", "turn19search0": "S19", "turn19search1": "S19", "turn19search2": "S20", "turn15search0": "S21" }, "sources": [ { "id": "S1", "authors": "Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, Percy Liang", "year": 2024, "title": "Lost in the Middle: How Language Models Use Long Contexts", "venue": "Transactions of the Association for Computational Linguistics 12", "url": "https://doi.org/10.1162/tacl_a_00638", "identifier": "doi:10.1162/tacl_a_00638", "status": "Published journal article", "check": "Publisher record and article identity verified; claim alignment checked against the reported multi-document QA and key-value retrieval findings.", "check_history": { "previous_scope": "Publisher record and article identity verified; claim alignment checked against the reported multi-document QA and key-value retrieval findings.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit.", "alternate_url": "https://aclanthology.org/2024.tacl-1.9/", "access_limit": "DOI inaccessible to checking tool; ACL publisher metadata/abstract read." } }, { "id": "S2", "authors": "Cheng-Ping Hsieh, Simeng Sun, Samuel Kriman, Shantanu Acharya, Dima Rekesh, Fei Jia, Yang Zhang, Boris Ginsburg", "year": 2024, "title": "RULER: What's the Real Context Size of Your Long-Context Language Models?", "venue": "COLM 2024", "url": "https://openreview.net/forum?id=kIoBbc76Sy", "identifier": "OpenReview: kIoBbc76Sy; arXiv:2404.06654", "status": "Conference paper", "check": "Conference paper and benchmark scope verified; claim alignment checked against task coverage and reported length degradation.", "check_history": { "previous_scope": "Conference paper and benchmark scope verified; claim alignment checked against task coverage and reported length degradation.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit.", "alternate_url": "https://arxiv.org/abs/2404.06654", "access_limit": "OpenReview gate; author preprint abstract read." } }, { "id": "S3", "authors": "Ali Modarressi, Hanieh Deilamsalehy, Franck Dernoncourt, Trung Bui, Ryan A. Rossi, Seunghyun Yoon, Hinrich Schuetze", "year": 2025, "title": "NoLiMa: Long-Context Evaluation Beyond Literal Matching", "venue": "Proceedings of Machine Learning Research 267, ICML 2025", "url": "https://proceedings.mlr.press/v267/modarressi25a.html", "identifier": "PMLR v267; arXiv:2502.05167", "status": "Peer-reviewed conference paper", "check": "Publisher record verified: 13 models were evaluated and 11 fell below half baseline at 32K. The supplied report says 12 and ten; corrected in this edition. Preserve the paper's baseline qualification.", "check_history": { "previous_scope": "Publisher record verified: 13 models were evaluated and 11 fell below half baseline at 32K. The supplied report says 12 and ten; corrected in this edition. Preserve the paper's baseline qualification.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S4", "authors": "Yushi Bai et al.", "year": 2024, "title": "LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding", "venue": "ACL 2024", "url": "https://aclanthology.org/2024.acl-long.172/", "identifier": "doi:10.18653/v1/2024.acl-long.172", "status": "Peer-reviewed conference paper", "check": "ACL publisher record and task taxonomy verified; claim used only for task heterogeneity, not as evidence of a universal context limit.", "check_history": { "previous_scope": "ACL publisher record and task taxonomy verified; claim used only for task heterogeneity, not as evidence of a universal context limit.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S5", "authors": "Jia He, Mukund Rungta, David Koleczek, Arshdeep Sekhon, Franklin X. Wang, Sadid Hasan", "year": 2024, "title": "Does Prompt Formatting Have Any Impact on LLM Performance?", "venue": "arXiv preprint", "url": "https://arxiv.org/abs/2411.10541", "identifier": "arXiv:2411.10541; doi:10.48550/arXiv.2411.10541", "status": "Preprint; no peer-reviewed venue identified in this check", "check": "arXiv metadata and abstract verified; the supplied ledger describes GPT-3.5/GPT-4 task-specific results. Do not label this a conference paper.", "check_history": { "previous_scope": "arXiv metadata and abstract verified; the supplied ledger describes GPT-3.5/GPT-4 task-specific results. Do not label this a conference paper.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S6", "authors": "Duc-Hai Nguyen, Vijayakumar Nanjappan, Barry O'Sullivan, Hoang D. Nguyen", "year": 2026, "title": "Questionnaire Meets LLM: A Benchmark and Empirical Study of Structural Skills for Understanding Questions and Responses", "venue": "LREC 2026", "url": "https://aclanthology.org/2026.lrec-1.371/", "identifier": "doi:10.63317/438xkvmy2xd9", "status": "Peer-reviewed conference paper", "check": "ACL record, study design, cross-format scope, and reported results verified.", "check_history": { "previous_scope": "ACL record, study design, cross-format scope, and reported results verified.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S7", "authors": "Yuan Sui, Mengyu Zhou, Mingjie Zhou, Shi Han, Dongmei Zhang", "year": 2024, "title": "Table Meets LLM: Can Large Language Models Understand Structured Table Data? A Benchmark and Empirical Study", "venue": "WSDM 2024", "url": "https://doi.org/10.1145/3616855.3635752", "identifier": "doi:10.1145/3616855.3635752", "status": "Peer-reviewed conference paper", "check": "ACM bibliographic record and representation comparisons verified; wording remains task-specific.", "check_history": { "previous_scope": "ACM bibliographic record and representation comparisons verified; wording remains task-specific.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit.", "alternate_url": "https://arxiv.org/abs/2305.13062", "access_limit": "DOI403; author preprint read." } }, { "id": "S8", "authors": "Masahiro Kato, Taka Kato", "year": 2026, "title": "Which Algorithm Specification Formats Help Language Models Implement Machine Learning Algorithms?", "venue": "arXiv preprint", "url": "https://arxiv.org/abs/2607.03158", "identifier": "arXiv:2607.03158", "status": "Preprint; no peer-reviewed venue identified in this check", "check": "Preprint identity and described controlled design checked. Findings must remain labeled preliminary and not treated as independently replicated.", "check_history": { "previous_scope": "Preprint identity and described controlled design checked. Findings must remain labeled preliminary and not treated as independently replicated.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S9", "authors": "Netanel Eliav", "year": 2026, "title": "Prompt Design at Scale: How Format, Instruction Count, and Context Length Shape Instruction Adherence and Hallucination in Large Language Models", "venue": "arXiv preprint", "url": "https://arxiv.org/abs/2607.19257", "identifier": "arXiv:2607.19257; VeyraBench project", "status": "Preprint; independent replication not identified", "check": "Preprint record, design, and reported format/context results checked. Results are from one study/corpus and should not be generalized beyond its conditions.", "check_history": { "previous_scope": "Preprint record, design, and reported format/context results checked. Results are from one study/corpus and should not be generalized beyond its conditions.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S10", "authors": "Zipeng Qiu, Chenyue Li, You Peng, Guangxin He, Binhang Yuan, Chen Wang", "year": 2026, "title": "TQA-Bench: Evaluating LLMs for Multi-Table Question Answering", "venue": "IEEE publication; arXiv version 2 dated 2026-06-05", "url": "https://arxiv.org/html/2411.19504", "identifier": "arXiv:2411.19504v2", "status": "2026 IEEE publication; arXiv full-text record", "check": "Current v2 identifies the 2026 IEEE publication. Full text directly reports tokenized lengths for Markdown, CSV, JSON, and HTML and serialization-performance comparisons. Claim basis confirmed; results remain specific to this table-QA benchmark.", "check_history": { "previous_scope": "Current v2 identifies the 2026 IEEE publication. Full text directly reports tokenized lengths for Markdown, CSV, JSON, and HTML and serialization-performance comparisons. Claim basis confirmed; results remain specific to this table-QA benchmark.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S11", "authors": "Tianyu Gao, Howard Yen, Jiatong Yu, Danqi Chen", "year": 2023, "title": "Enabling Large Language Models to Generate Text with Citations", "venue": "EMNLP 2023", "url": "https://aclanthology.org/2023.emnlp-main.398/", "identifier": "doi:10.18653/v1/2023.emnlp-main.398", "status": "Peer-reviewed conference paper", "check": "ACL record and separation of fluency, correctness, and citation-quality evaluation verified.", "check_history": { "previous_scope": "ACL record and separation of fluency, correctness, and citation-quality evaluation verified.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S12", "authors": "Cheng Niu et al.", "year": 2024, "title": "RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Models", "venue": "ACL 2024", "url": "https://aclanthology.org/2024.acl-long.585/", "identifier": "doi:10.18653/v1/2024.acl-long.585", "status": "Peer-reviewed conference paper", "check": "ACL record, corpus size phrasing ('nearly 18,000'), and annotation scope verified.", "check_history": { "previous_scope": "ACL record, corpus size phrasing ('nearly 18,000'), and annotation scope verified.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S13", "authors": "Sewon Min et al.", "year": 2023, "title": "FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation", "venue": "EMNLP 2023", "url": "https://aclanthology.org/2023.emnlp-main.741/", "identifier": "doi:10.18653/v1/2023.emnlp-main.741", "status": "Peer-reviewed conference paper", "check": "ACL record and atomic-fact evaluation method verified.", "check_history": { "previous_scope": "ACL record and atomic-fact evaluation method verified.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S14", "authors": "Huiqiang Jiang, Qianhui Wu, Chin-Yew Lin, Yuqing Yang, Lili Qiu", "year": 2023, "title": "LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models", "venue": "EMNLP 2023", "url": "https://aclanthology.org/2023.emnlp-main.825/", "identifier": "doi:10.18653/v1/2023.emnlp-main.825", "status": "Peer-reviewed conference paper", "check": "ACL record and prompt-compression method/results verified; does not establish preservation of all evidence properties.", "check_history": { "previous_scope": "ACL record and prompt-compression method/results verified; does not establish preservation of all evidence properties.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S15", "authors": "Huiqiang Jiang et al.", "year": 2024, "title": "LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression", "venue": "ACL 2024", "url": "https://aclanthology.org/2024.acl-long.91/", "identifier": "doi:10.18653/v1/2024.acl-long.91", "status": "Peer-reviewed conference paper", "check": "ACL record and reported NaturalQuestions example verified; result is task/model-specific.", "check_history": { "previous_scope": "ACL record and reported NaturalQuestions example verified; result is task/model-specific.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S16", "authors": "Fangyuan Xu, Weijia Shi, Eunsol Choi", "year": 2024, "title": "RECOMP: Improving Retrieval-Augmented LMs with Context Compression and Selective Augmentation", "venue": "ICLR 2024", "url": "https://proceedings.iclr.cc/paper_files/paper/2024/hash/bda88ed2892f5e61c9a9bf215c566913-Abstract-Conference.html", "identifier": "ICLR 2024 proceedings record", "status": "Peer-reviewed conference paper", "check": "Authors and selective context-compression scope checked against ICLR proceedings, 2026-10-04; earlier Vivek Sharma attribution corrected.", "check_history": { "previous_scope": "Authors and selective context-compression scope checked against ICLR proceedings, 2026-10-04; earlier Vivek Sharma attribution corrected.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S17", "authors": "Jeff Wu, Long Ouyang, Daniel M. Ziegler, Nisan Stiennon, Ryan Lowe, Jan Leike, Paul Christiano", "year": 2021, "title": "Recursively Summarizing Books with Human Feedback", "venue": "arXiv preprint and OpenAI research publication", "url": "https://arxiv.org/abs/2109.10862", "identifier": "arXiv:2109.10862", "status": "Preprint/research report; not represented here as a peer-reviewed venue paper", "check": "Primary report and method checked; it supports specialized recursive summarization, not lossless evidence preservation.", "check_history": { "previous_scope": "Primary report and method checked; it supports specialized recursive summarization, not lossless evidence preservation.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S18", "authors": "Han Wang, Archiki Prasad, Elias Stengel-Eskin, Mohit Bansal", "year": 2025, "title": "AdaCAD: Adaptively Decoding to Balance Conflicts between Contextual and Parametric Knowledge", "venue": "NAACL 2025", "url": "https://aclanthology.org/2025.naacl-long.581/", "identifier": "ACL Anthology: 2025.naacl-long.581; arXiv:2409.07394", "status": "Peer-reviewed conference paper", "check": "Updated publication record and context/parametric-conflict scope verified; use 2025 venue citation rather than treating it solely as a 2024 preprint.", "check_history": { "previous_scope": "Updated publication record and context/parametric-conflict scope verified; use 2025 venue citation rather than treating it solely as a 2024 preprint.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S19", "authors": "Lianmin Zheng et al.", "year": 2023, "title": "Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena", "venue": "NeurIPS 2023 Datasets and Benchmarks Track", "url": "https://papers.neurips.cc/paper_files/paper/2023/hash/91f18a1287b398d378ef22505bf41832-Abstract-Datasets_and_Benchmarks.html", "identifier": "NeurIPS 2023 paper record; arXiv:2306.05685", "status": "Peer-reviewed conference paper", "check": "NeurIPS record and stated judge-bias analysis verified; reported limitations retained.", "check_history": { "previous_scope": "NeurIPS record and stated judge-bias analysis verified; reported limitations retained.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S20", "authors": "Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, Chenguang Zhu", "year": 2023, "title": "G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment", "venue": "EMNLP 2023", "url": "https://aclanthology.org/2023.emnlp-main.153/", "identifier": "doi:10.18653/v1/2023.emnlp-main.153", "status": "Peer-reviewed conference paper", "check": "ACL record, reported 0.514 Spearman correlation on summarization, and bias caveat verified.", "check_history": { "previous_scope": "ACL record, reported 0.514 Spearman correlation on summarization, and bias caveat verified.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } }, { "id": "S21", "authors": "Taylor Berg-Kirkpatrick, David Burkett, Dan Klein", "year": 2012, "title": "An Empirical Investigation of Statistical Significance in NLP", "venue": "EMNLP-CoNLL 2012", "url": "https://aclanthology.org/D12-1091/", "identifier": "ACL Anthology: D12-1091", "status": "Peer-reviewed conference paper", "check": "ACL record verified. Its methodological relevance is general; it does not validate this dossier's proposed particular analysis plan.", "check_history": { "previous_scope": "ACL record verified. Its methodological relevance is general; it does not validate this dossier's proposed particular analysis plan.", "october4_scope": "Primary metadata/abstract recheck; not a full experiment, numerical or quotation audit." } } ], "claims": [ { "id": "C01", "sources": [ "S5", "S6", "S7", "S8", "S9", "S10" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C02", "sources": [ "S5", "S6", "S8", "S9" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C03", "sources": [ "S5", "S6", "S7", "S8", "S9", "S10" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C04", "sources": [ "S1", "S2", "S3" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C05", "sources": [ "S2", "S3", "S4" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C06", "sources": [ "S9" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C07", "sources": [ "S9", "S10" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C08", "sources": [ "S14", "S15", "S16" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C09", "sources": [ "S14", "S15", "S16" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C10", "sources": [ "S2", "S3" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C11", "sources": [ "S13" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C12", "sources": [ "S12" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C13", "sources": [ "S11" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C14", "sources": [], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C15", "sources": [ "S11" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C16", "sources": [ "S7", "S8", "S9", "S10" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C17", "sources": [ "S8", "S9", "S10" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C18", "sources": [ "S17" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C19", "sources": [ "S17" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C20", "sources": [ "S14", "S15", "S16", "S17" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C21", "sources": [ "S19", "S20" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C22", "sources": [ "S18" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C23", "sources": [ "S9", "S18" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C24", "sources": [ "S9", "S10", "S14", "S15", "S16" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C25", "sources": [], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." }, { "id": "C26", "sources": [ "S2", "S3", "S11", "S12", "S13" ], "check_scope": "Inherited source-pointer map; editorial scope revised, not full claim verification." } ], "transformation_record": [ "Original source digest retained; original public payload withdrawn under owner direction.", "Revised claim scope and replication labels; corrected S16 author attribution and NoLiMa absolute/relative baseline wording.", "Stable source/claim IDs and historical crosswalk retained; no complete claim or quotation audit." ], "resource_version": "1.4", "provenance": { "tool_provider": "OpenAI", "tool": "OpenAI Deep Research", "resolved_model": "gpt-5-thinking", "model_reasoning_generation_metadata": "v5", "research_completed": "2026-09-24", "author_attribution": "OpenAI Deep Research, using GPT-5 Thinking, as supplied by the human facilitator; identity not independently verified.", "submitter": "Human facilitator", "arc_editor": "Lumen (GPT-6.1 Sol)", "reviewer": "No independent reviewer", "review_scope": "Constitutional editorial review and selected primary-record checks; original study-detail checks remain inherited where not specifically rechecked. No complete claim/quotation audit." }, "constitutional_revision": { "version": "1.4", "source_report_withdrawn": true, "report_alias_is_current_revision": true, "claim_label_definition": "Cross-study reports are related reported findings; exact replication or independent lineages not established.", "additional_counterevidence": "https://aclanthology.org/2024.acl-long.736/" } }