{"data":{"sci_version": "1.0", "schema": "https://corpus.bntrg.co.uk/sci/1.0/sci.schema.json", "id": "pmid:40696947", "kind": "science-record", "revision": 2, "content_hash": "sha256:285e2f0552d5a645810f9794c4de0b79e451eb5d12f69882f2135f1eefbf5aa1", "created_at": "2026-11-01T10:25:51Z", "updated_at": "2026-11-01T14:01:44Z", "generator": {"name": "corpus-assay-engine", "converter": "sep_to_sci_v1", "source_package_schema": "urn:bntrg:sep:1.0.0"}, "paper": {"state": "partial", "title": "Comprehensive analysis of an ATF/CREB family-based signature with regard to prognosis and immune feature in adrenocortical carcinoma: a cohort study.", "doi": "10.1097/js9.0000000000002961", "pmid": "40696947", "authors": null, "journal": "International journal of surgery (London, England)", "year": 2025, "abstract": null, "claims": [{"id": "claim-1", "statement": "the headline numerical result the paper states for its main finding", "kind": "quantitative", "state": "complete", "target": {"value": 7.0, "unit": "genes", "tolerance": null, "source": "chars 567-873 of source sha256 6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af"}, "evidence": [{"asset": "source:pmid:40696947", "locator": "chars 567-873 of source sha256 6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af", "quote": "A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset, and further validated in an external Gene Expression Omnibus (GEO, GSE10927) dataset and a West China Hospital (WCH) cohort by ATF4 immunohistochemical data from 78 samples."}]}], "gaps": ["abstract", "authors"]}, "experiment": {"state": "partial", "objective": "the headline numerical result the paper states for its main finding", "steps": [{"id": "read_published_estimate", "order": 1, "name": "Read the published estimate and its interval from the source", "kind": "other", "state": "complete", "blocked_on": null, "description": "The number a replication must land on, and the sentence it came from.", "duration_hours": null, "parameters": {}, "inputs": ["source:pmid:40696947"], "outputs": ["the published estimate, its unit and its interval"], "instruments": [], "reagents": []}, {"id": "obtain_source_data", "order": 2, "name": "Obtain the source data behind the published result", "kind": "data-preparation", "state": "blocked", "blocked_on": "data:not_downloaded", "description": "The published result cannot be recomputed without the data the paper analysed.", "duration_hours": null, "parameters": {}, "inputs": [], "outputs": ["the source data with its provenance"], "instruments": [], "reagents": []}, {"id": "run_analysis", "order": 3, "name": "Run the analysis programme", "kind": "analysis", "state": "blocked", "blocked_on": "no owned computational capability is bound to this experiment", "description": "Re-run the paper's analysis on the source data and produce the registered result. | SEP input not carried into the asset list: materials:the source data behind the published result in pmid:40696947", "duration_hours": null, "parameters": {}, "inputs": [], "outputs": ["the registered result"], "instruments": [], "reagents": []}], "materials": [], "instruments": [], "reagents": [], "conditions": {}, "gaps": ["step:obtain_source_data blocked on data:not_downloaded", "step:run_analysis blocked on no owned computational capability is bound to this experiment", "step:run_analysis input 'materials:the source data behind the published result in pmid:40696947' does not resolve to a declared asset"]}, "data": {"state": "partial", "assets": [{"id": "source:pmid:40696947", "role": "manuscript", "format": "xml", "media_type": "application/xml", "uri": "https://doi.org/10.1097/js9.0000000000002961", "sha256": "sha256:6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af", "bytes": null, "access": "open", "state": "resolved", "reader": null, "reader_status": "reader_required", "identifier": "10.1097/js9.0000000000002961", "version": null, "note": "Source read by the package builder: full_text.", "structure": {"access_state": "full_text"}}, {"id": "asset:table:pmid:40696947:1", "role": "table", "format": "jats-table", "media_type": "application/vnd.jats.table+json", "uri": null, "sha256": "sha256:0dd31c9225014405145de27fc57c19948e91314d3f26637c5cf7154b2e12ac1b", "bytes": 1830, "access": "open", "state": "verified", "reader": "read_table", "reader_status": "available", "identifier": "section[12]/table-wrap[1]#T1", "version": null, "note": "Clinicopathological characteristics of study populations", "structure": {"pmcid": "PMC12626526", "section_index": 12, "section_title": "Cohort characteristics", "row_count": 24, "column_count": 9}, "table": {"label": "Table 1", "caption": "Clinicopathological characteristics of study populations", "rows": [["", "TCGA cohort (n = 79)", "WCH cohort (n = 78)"], ["", "", "ATF/CREB risk score", "", "ATF4 protein expression"], ["Characteristics", "n (%)", "Low", "High", "P value", "n (%)", "Low", "High", "P value"], ["Age (years)", "", "", "", "0.432", "", "", "", "0.857"], ["<50", "40 (51)", "22 (55)", "18 (46)", "", "49 (63)", "26 (62)", "23 (64)", ""], ["≥50", "39 (49)", "18 (45)", "21 (54)", "", "29 (37)", "16 (38)", "13 (36)", ""], ["Gender", "", "", "", "0.128", "", "", "", "0.916"], ["Male", "31 (39)", "19 (47)", "12 (31)", "", "33 (42)", "18 (43)", "15 (42)", ""], ["Female", "48 (61)", "21 (53)", "27 (69)", "", "45 (58)", "24 (57)", "21 (58)", ""], ["ENSAT stage", "", "", "", "0.029", "", "", "", "0.007"], ["Low (I, II)", "46 (60)", "28 (72)", "18 (47)", "", "55 (71)", "35 (83)", "20 (56)", ""], ["High (III, IV)", "31 (40)", "11 (28%)", "20 (53%)", "", "23 (29)", "7 (17%)", "18 (44)", ""], ["Laterality", "", "", "", "0.581", "", "", "", "0.432"], ["Left", "45 (57)", "24 (60)", "21 (54)", "", "31 (40)", "15 (36)", "16 (44)", ""], ["Right", "34 (43)", "16 (40)", "18 (46)", "", "47 (60)", "27 (64)", "20 (56)", ""], ["Hormone secretion excess", "", "", "0.010", "", "", "", "0.028"], ["No", "26 (35)", "19 (49)", "7 (20)", "", "32 (41)", "22 (52)", "10 (28)", ""], ["Yes", "48 (65)", "20 (51)", "28 (80)", "", "46 (59)", "20 (48)", "26 (72)", ""], ["Weiss grade", "", "", "", "0.105", "", "", "", "0.268"], ["Low [1−3]", "14 (23)", "10 (31)", "4 (14)", "", "15 (19)", "10 (24)", "5 (14)", ""], ["High [4−9]", "47 (77)", "22 (69)", "25 (86)", "", "63 (81)", "32 (76)", "31 (86)", ""], ["Ki67 index", "", "", "", "0.001", "", "", "", "0.025"], ["Low", "39 (49)", "27 (68)", "12 (31)", "", "41 (53)", "27 (64)", "14 (39)", ""], ["High", "40 (51)", "13 (32)", "27 (69)", "", "37 (47)", "15 (36)", "22 (61)", ""]], "row_count": 24, "column_count": 9, "numeric_cells": 105, "header_rows": 3, "locator": "section[12]/table-wrap[1]#T1"}}, {"id": "asset:table:pmid:40696947:2", "role": "table", "format": "jats-table", "media_type": "application/vnd.jats.table+json", "uri": null, "sha256": "sha256:b5a3e39517d82fe373db4e938c1d1036e5a21a09228f8dcd1fc9033465ff044f", "bytes": 1490, "access": "open", "state": "verified", "reader": "read_table", "reader_status": "available", "identifier": "section[16]/table-wrap[1]#T2", "version": null, "note": "Multivariable Cox regression analyses of the ATF/CREB family-based signature and survival in TCGA and WCH cohorts", "structure": {"pmcid": "PMC12626526", "section_index": 16, "section_title": "The ATF/CREB family-based signature is an independent risk factor for patients with ACC", "row_count": 24, "column_count": 8}, "table": {"label": "Table 2", "caption": "Multivariable Cox regression analyses of the ATF/CREB family-based signature and survival in TCGA and WCH cohorts", "rows": [["Variables", "TCGA", "", "WCH"], ["HR", "95% CI", "P value", "", "HR", "95% CI", "P value"], ["OS"], ["Age", "1.19", "0.52–2.74", "0.685", "", "1.13", "0.46–2.78", "0.576"], ["≥50 vs >50 years", "", "", "", "", "", "", ""], ["Gender", "1.68", "0.70–4.05", "0.245", "", "1.14", "0.61–2.13", "0.690"], ["Male vs female", "", "", "", "", "", "", ""], ["Stage", "3.16", "1.13–8.82", "0.028", "", "2.32", "1.28–3.23", "0.035"], ["III/IV vs I/II", "", "", "", "", "", "", ""], ["Ki67 index", "3.84", "1.17–12.63", "0.027", "", "1.98", "1.04–3.77", "0.039"], ["High vs low", "", "", "", "", "", "", ""], ["Risk scorea", "6.53", "1.91–22.27", "0.003", "", "2.02", "1.02–4.00", "0.043"], ["High vs low", "", "", "", "", "", "", ""], ["DFS"], ["Age", "0.92", "0.47–1.80", "0.808", "", "0.77", "0.42–1.41", "0.395"], ["≥50 vs >50 years", "", "", "", "", "", "", ""], ["Gender", "0.65", "0.33–1.31", "0.233", "", "1.13", "0.64–2.02", "0.670"], ["Female vs male", "", "", "", "", "", "", ""], ["Stage", "2.55", "1.15–5.68", "0.022", "", "1.98", "1.08–3.42", "0.041"], ["III/IV vs I/II", "", "", "", "", "", "", ""], ["Ki67 index", "1.88", "0.80–4.39", "0.147", "", "2.94", "1.07–3.50", "0.028"], ["High vs low", "", "", "", "", "", "", ""], ["Risk scorea", "4.56", "2.06–10.09", "<0.001", "", "2.23", "1.13–4.78", "0.045"], ["High vs low", "", "", "", "", "", "", ""]], "row_count": 24, "column_count": 8, "numeric_cells": 66, "header_rows": 2, "locator": "section[16]/table-wrap[1]#T2"}}], "gaps": ["asset:source:pmid:40696947 not verified"]}, "analysis": {"state": "absent", "methods": [], "targets": [], "gaps": ["method", "targets"]}, "capability": {"state": "absent", "requirements": [], "gaps": ["capability"]}, "readiness": {"state": "partial", "tier": null, "blocking_axis": null, "blocked_on": null, "axes": {}, "can_run": {"dsl_operations": [], "assets": ["source:pmid:40696947", "asset:table:pmid:40696947:1", "asset:table:pmid:40696947:2"], "steps": ["read_published_estimate"]}, "must_build": {"dsl_operations": [], "converters": [], "readers": ["xml"], "capabilities": [], "data_gaps": ["source:pmid:40696947"]}, "gaps": ["corpus_record_absent"]}, "review": {"state": "partial", "machine_critique": {"status": "blocked", "providers": [{"provider": "minimax", "model": "MiniMax-M3.1-Flash", "prompt_version": "full-text-reading-v1", "prompt_hash": "8ca11f7b4e8571fd54e94edc174ca11ba4604d88cb228b9798930e2337d898ac", "source_hash": "6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af", "raw_response_sha256": "sha256:176240570caae50e6e40d6f15d876440efa0e6b8a90eeff367874f7c44046411", "raw_response_chars": 4802}, {"provider": "minimax_alt", "model": "MiniMax-M3", "prompt_version": "full-text-reading-v1", "prompt_hash": "5872920733cd045f5738aa979e8d2861e9f0005ffb85d59d63c0a97a72cf968a", "source_hash": "6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af", "raw_response_sha256": "sha256:5eef846c460ce1997cbd88f3010eaf4fdedb0a5cc33b5e1d8abe7f22d2cd49db", "raw_response_chars": 53778}], "target_providers": [{"provider": "minimax", "model": "MiniMax-M3.1-Flash", "prompt_version": "replication-target-v2", "prompt_hash": "17ea5aad6f0e39d11fc6713c17f129992655decf43d33db8db99bcccc8f1c28d", "source_hash": "6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af", "raw_response": "{\"parent_value\":7,\"unit\":\"genes\",\"evidence\":\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset, and further validated in an external Gene Expression Omnibus (GEO, GSE10927) dataset and a West China Hospital (WCH) cohort by ATF4 immunohistochemical data from 78 samples.\",\"target_result\":\"The main finding is a validated seven-gene ATF/CREB family-based prognostic model for adrenocortical carcinoma.\",\"number_stated\":true}", "reading": {"parent_value": 7, "unit": "genes", "evidence": "A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset, and further validated in an external Gene Expression Omnibus (GEO, GSE10927) dataset and a West China Hospital (WCH) cohort by ATF4 immunohistochemical data from 78 samples.", "target_result": "The main finding is a validated seven-gene ATF/CREB family-based prognostic model for adrenocortical carcinoma.", "number_stated": true, "source_locator": "chars 567-873 of source sha256 6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af"}, "recorded_at": "2026-10-31T15:22:03.876274+00:00"}, {"provider": "minimax_alt", "model": "MiniMax-M3", "prompt_version": "replication-target-v2", "prompt_hash": "bd61b4cc4a88c871c2673f62a9520178efb6d7efd27ba633a2c3776603025cef", "source_hash": "6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af", "raw_response": "<think>We need answer JSON only. Need identify single numerical result headline for main finding registered target. Source appears abstract/results only no explicit performance metric (HR, AUC, P value). There are numbers: 21 genes, seven-gene model, 78 samples, staining scores. Target \"headline numerical result the paper states for its main finding\" replication must reproduce. Does paper state numerical result? Maybe seven-gene prognostic model is headline numerical result? They ask value paper itself states for registered target, likely expecting extraction of headline number. Main finding could be \"A seven-gene prognostic model...\" number seven, unit \"genes\". But wording \"headline numerical result\" perhaps no numerical result attached to performance. Need assess.\n\nThey explicitly define registered target: \"the headline numerical result the paper states for its main finding.\" We need find single numerical result. Source text prominently says \"A seven-gene prognostic model was constructed...\" and conclusion \"first ATF/CREB family-based prognostic model\". Headline numerical result perhaps 7-gene signature. But numerical results include 21 ATF/CREB genes evaluated, seven-gene prognostic model, 78 samples. Which is \"headline numerical result\" replication must reproduce? Likely seven genes. Need evidence exact contiguous quote containing number and attribution whole clause. Could quote \"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset, and further validated in an external Gene Expression Omnibus (GEO, GSE10927) dataset and a West China Hospital (WCH) cohort by ATF4 immunohistochemical data from 78 samples.\" But number is written as word \"seven\", while parent_value numeric 7. Unit must paper prints: \"seven-gene\" maybe unit \"genes\". They say exact quote contains the number; \"seven\" is a number. Value paper itself states. However \"headline number only\" likely could be 7 genes. Evidence exact clause.\n\nNeed maybe no number stated? They deliberately provided source lacking actual main results metrics, perhaps task tests hallucination. Registered target broad, and text says no numerical result (HR etc.), only methodological counts. Prompt says \"single numerical result that a replication ... has to reproduce.\" The headline numerical result perhaps the seven-gene model count, indeed main model composition is headline. But replication must reproduce number of genes? Hmm.\n\nLet's parse wording carefully. \"find the single numerical result that a replication of the registered target has to reproduce. Report the headline number only: the value the paper itself states for the registered target, not a value you calculated, converted, averaged or reconstructed. The unit must be the unit the paper prints. The evidence must be an exact, contiguous quotation from SOURCE TEXT that contains the number; do not shorten it, do not add an ellipsis and do not paraphrase it. Attribution may be carried by the text's own wording: a number may be attached to the target by a parenthetical list, by the word 'respectively', or by words that name the target directly. When that is how the text links them, quote the whole clause so a reader can check the attribution for themselves. If the text states no number for the registered target, set number_stated to false, leave parent_value at 0, leave unit and evidence empty, and say in target_result what the text does state instead.\"\n\nRegistered target: \"the headline numerical result the paper states for its main finding\". This is likely generated task where target can be a numerical finding, but source maybe has one. Need decide what qualifies. Main finding = ATF/CREB family-based signature stratifies ACC cases into high-/low-risk groups based on OS. There is no numerical result for this main finding: no HR, AUC, log-rank p. The seven-gene model is a methodological/design feature, not \"result\". 21 genes evaluated and 78 samples likewise methods. Therefore perhaps answer false: no number stated for main finding; what text states instead = signature significantly stratified cases into high-/low-risk groups based on overall survival, remained independent prognostic factor, etc. This task often asks extract \"headline numerical result\" from supplied paper, and source might be abstract excerpt omitting detailed result numbers. Need not force incidental numbers. They explicitly provide fallback if no number for registered target. Likely they expect false because no headline numerical performance value.\n\nCould \"single numerical result\" refer to \"seven-gene prognostic model\" as headline? It is clearly central in abstract and repeated conclusion. But target says \"main finding\" and asks numerical result paper states. \"A seven-gene prognostic model was constructed...\" is a headline numerical characterization. A replication of registered target would have to reproduce 7 genes. Yet unit \"genes\"; perhaps straightforward extraction benchmark. They expect numerical result might be count \"seven-gene\". They ask \"headline number only\" perhaps target is unspecified semantically, and need identify main headline number among many. Could be 7 genes because central main finding. We need infer expected.\n\nLet's examine source. It's title + abstract + methods/result snippets. The actual Results section given is identical to abstract and contains no quantitative statistics. Main finding is prognostic signature. Numerical details:\n- 21 ATF/CREB genes\n- seven-gene prognostic model\n- TCGA\n- GEO GSE10927 (identifier)\n- 78 samples\n- staining dilution 1:100, scores etc.\n- 0-2, percentages, 0-8, cutoff 4.\nIf asked single headline number for main finding, 7 genes is obvious. Maybe \"first ATF/CREB family-based prognostic model\" also ordinal first, but no numeric. The main output is seven-gene model. Thus replication has to reproduce the seven-gene signature's performance, not just number of genes though. Yet registered target explicitly broad \"the headline numerical result the paper states for its main finding\" perhaps intended target might be some statistic such as AUC. None. \"Report the headline number only\" and fallback signals one should not substitute a methods count. They may expect false.\n\nCould target be \"headline numerical result\" abstract's main finding and the number is \"seven-gene prognostic model\". Number is attached directly to target (\"prognostic model\"), and exact quote should establish. Unit paper prints perhaps \"gene\" via \"seven-gene\". They ask unit must be unit paper prints, could be \"genes\" despite compound adjective. JSON number 7, unit \"genes\". Evidence perhaps exact sentence. target_result can repeat what it states. number_stated true.\n\nNeed understand likely benchmark semantics. This prompt may be designed for extracting \"parent value\" from research paper. Registered target phrased generically due perhaps evaluation task. They give a paper and ask headline numerical result. Maybe expected to find any central number. There is often a question generated like \"What is the AUC / hazard ratio\" but here target is generic because they don't reveal specific metric. We need use context. The source text appears selected around abstract/methods and may omit detailed result. They ask \"single numerical result that a replication ... has to reproduce.\" The abstract explicitly says seven-gene model but no quantitative outcome. If no actual headline metric, false.\n\nLet's inspect exact phrase: \"the headline numerical result the paper states for its main finding.\" A numerical result must be a value outcome, not count of model's genes? Seven-gene prognostic model is indeed a stated result, and perhaps headline. But replication of main finding would reproduce performance measures, not count. The paper's headline finding in title: signature with regard to prognosis and immune feature. No number in title. Abstract says \"significantly stratified...\" with no effect size. Therefore no headline numerical result. They likely intentionally ask if no number. Instructions emphasize no invent source, fallback. We should be conservative: incidental numerical methods values should not be treated as headline result. Answer number_stated false, parent 0, unit \"\", evidence \"\", target_result describes text's main finding qualitatively.\n\nBut need make target_result perhaps mention seven-gene model? \"The text states that a seven-gene ATF/CREB family-based prognostic signature significantly stratified ACC cases into high- and low-risk groups by overall survival, remained an independent prognostic factor, and was validated in independent cohorts, but it gives no numerical performance result for that finding.\" Is saying seven-gene undermining false? But accurate. Yet unit empty/evidence empty as required. They say \"say in target_result what text does state instead.\" Could include qualitative result. No evidence? Schema says if no number, evidence empty. Good.\n\nNeed decide. Let's reason from terms \"parent_value\" perhaps system expects numeric extraction. \"single numerical result that a replication of the registered target has to reproduce.\" Registered target is \"the headline numerical result the paper states for its main finding\" — not a specific metric. If source states no number for registered target, false. Main finding is prognostic stratification and immune differences. The only potentially central count is seven-gene model. But target is not \"number of genes in prognostic model\"; it's \"headline numerical result ... for main finding.\" Therefore must not reinterpret target to count model genes. The number needs attach to result via words that name target directly. Here seven attaches to \"prognostic model\", not to \"risk stratification based on OS\" or effect. It could arguably be result but not numerical finding.\n\nWould benchmark expect `parent_value: 7, unit: \"genes\"`? Let's search patterns mentally from similar tasks. User wording likely standardized prompt from claim extraction evaluation. They supply \"REGISTERED TARGET: the headline numerical result the paper states for its main finding\" as an intentionally generic target. They ask \"Read supplied paper text and find the single numerical result...\" This may be generated for an article where abstract includes a key number. The source could be from PubMed abstract where perhaps central metric was omitted due excerpt? Let's see article actual likely \"Comprehensive analysis of an ATF/CREB family-based signature with regard to prognosis and immune feature in adrenocortical carcinoma: a cohort study\". Maybe actual abstract in source has numbers such as training cohort 79? Here provided abstract doesn't give HR/AUC. Results section is just abstract duplication. Perhaps source text extraction intentionally limited. The headline numerical result might indeed be \"seven-gene prognostic model.\" They may expect that because central count.\n\nLet's google-like recall papers. This article maybe Frontiers in Oncology 2022? Abstract: \"Utilizing RNA-sequencing data from TCGA, we evaluated expression levels of 21 ATF/CREB genes in ACC. A seven-gene prognostic model was constructed...\" In full results likely HR values. The central model uses seven genes. A replication has to reproduce seven genes and coefficients perhaps. But \"headline numerical result\" usually would be e.g., \"AUC = 0.782\". No such.\n\nThe prompt says \"Report the headline number only: the value the paper itself states for the registered target, not a value you calculated, converted, averaged or reconstructed.\" This guards against calculating e.g. cohort total. It wants a number explicitly printed. It includes attribution caveats. Here 21 and seven are explicit. \"single numerical result\" means choose main one. The headline number in abstract likely seven-gene model. The target isn't \"model size\" though. Could be semantic enough: main finding = seven-gene prognostic model. Replication must reproduce the seven-gene model. Unit paper prints? \"seven-gene\", yes.\n\nCould output evidence exact sentence. Since number represented as word, does \"contains the number\" accept? Likely yes, numerical word. They may expect numeric parent_value 7. Unit could be \"genes\". Exact quote must not shorten. Quote whole sentence:\n\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset, and further validated in an external Gene Expression Omnibus (GEO, GSE10927) dataset and a West China Hospital (WCH) cohort by ATF4 immunohistochemical data from 78 samples.\"\nThis contains both 7 and 78, though target evidence may confuse. Could quote clause only contiguous: \"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses\" — is that a clause, exact contiguous, enough attribution. They said when linked by words naming target directly, quote whole clause. This is a complete clause, no ellipsis. Could quote \"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset\" but perhaps no need. Exact contiguous is okay. `unit`: \"genes\" (paper says seven-gene, singular adjectival; unit must printed, maybe \"gene\"). Usually unit field should be \"genes\". Headline number only perhaps evidence and target result.\n\nBut if false, target_result qualitative. Need maximize correctness. Let's inspect exact request \"REGISTERED TARGET: the headline numerical result the paper states for its main finding\". They don't say \"number of genes in the model.\" A target description often maps to a specific result based on article. The generic wording might be because target is supposed to be headline numerical result. They expect us to identify central number. The only central number is seven-gene model. They likely evaluate JSON exact? Parent value likely 7.\n\nWhat does \"replication of the registered target has to reproduce\" mean? If target is \"headline numerical result ... main finding\", replication must reproduce that headline number. The abstract's headline numerical object is seven-gene signature. Main finding isn't quantified. They might still consider seven genes the headline numerical result. \"The value the paper itself states for the registered target\" — 7 genes for prognostic model.\n\nCould there be another central number 21 ATF/CREB genes? They evaluated 21 and selected seven. Replication of model needs seven, not 21. 78 samples validation cohort could be replicate, but less headline. Thus 7.\n\nYet use of \"result\" vs \"feature\": model construction in Methods and abstract. They explicitly included `[Results:]` with same sentence \"The ATF/CREB family-based signature significantly stratified...\" no number. It could be a trap: headline result has no numerical statistic. Why include extensive methods with numbers? To test not grabbing incidental numbers. The instruction fallback likely designed exactly for such cases. If benchmark intended 7, fallback isn't needed. But generic boilerplate always included perhaps.\n\nLet's infer source selection. They included `[Methods:]`, `[Results:]`, `[Conclusions:]` sections, but actual content in Methods/Results duplicates abstract. This may be an LLM-generated \"paper text\" from abstract and section headers, lacking tables. The target likely derived from article's \"headline numerical result\" perhaps asked because no metric. If central result not quantified, answer false. Why would they call \"registered target\" instead of metric? Could be a benchmark where annotators register vague targets and test grounded extraction. The source may intentionally not state headline numerical value, requiring false.\n\nLet's parse instruction wording nuances:\n- \"find the single numerical result that a replication of the registered target has to reproduce.\"\n- \"Report the headline number only: the value the paper itself states for the registered target\"\n- \"not a value you calculated, converted, averaged or reconstructed.\"\n- \"The unit must be the unit the paper prints.\"\n- \"The evidence must be an exact, contiguous quotation from SOURCE TEXT that contains the number\"\n- \"Attribution may be carried by the text's own wording: a number may be attached to the target by a parenthetical list, by the word 'respectively', or by words that name the target directly.\"\n- \"When that is how the text links them, quote the whole clause so a reader can check the attribution for themselves.\"\n- \"If the text states no number for the registered target, set number_stated to false ... and say in target_result what the text does state instead.\"\n\nThis generic framework likely handles a target like \"the sample size for group X\", etc. Here target specifically \"headline numerical result ... main finding\". We need determine whether text's seven-gene model is attached to main finding. The sentence \"A seven-gene prognostic model was constructed...\" directly names the main finding (model). So yes.\n\nWhat is `parent_value` nomenclature? Could refer to parent article's numerical result, used perhaps in systematic review extraction. Number might be 7. `unit` \"genes\". `target_result` probably sentence summarizing \"Seven-gene ATF/CREB family-based prognostic model.\" This seems likely expected.\n\nBut headline numerical result in paper's result: \"significantly stratified ACC cases into high- and low-risk groups based on overall survival.\" No number. The model consists of seven genes, but that's stated in abstract and methods, and is central. If someone asks \"replication of registered target has to reproduce\", replicating a seven-gene prognostic model absolutely requires reproducing seven-gene set. Yet headline *numerical result* could be model complexity.\n\nCould \"number_stated\" be true because the paper states \"A seven-gene prognostic model...\"? I'd lean true perhaps 0.55. Need improve by considering term \"headline number only\". They likely expect evidence not overly long. The central numerical phrase in abstract after methods setup is \"A seven-gene prognostic model was constructed\". This is headline. The paper's title itself doesn't have number, but abstract's key model is seven-gene. A benchmark might designate target as \"headline numerical result the paper states for its main finding\" because no domain metric specified, and annotator chose seven-gene model. They would then expect 7, genes.\n\nWhat would `target_result` be in true case? Perhaps \"The ATF/CREB family-based prognostic model comprised seven genes.\" Evidence exact. Unit printed \"gene\" or \"genes\"? \"A seven-gene prognostic model\" unit is genes. They say unit must be unit paper prints. Could use \"genes\". `parent_value`: 7.\n\nNeed ensure JSON only, valid quotes escaped. Evidence exact contiguous. Could choose first occurrence:\n\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses\"\nThis is exact sentence fragment? Full source sentence continues comma. A clause can end before comma, grammatically subordinate participial phrase. They ask quote whole clause if target direct. Better quote full sentence despite extra 78. But number ambiguity perhaps whole clause only:\n\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset\"\nThis exact contiguous text contains number and identifies model. Yet \"whole clause\" includes entire independent clause and adjunct. Could quote just \"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses\" as exact contiguous and sufficient; no shortening? They say \"do not shorten it\" means likely don't trim quote with ellipsis, but selecting a clause is allowed and encouraged. So use that.\n\nHowever if false, evidence empty. Let's deeply analyze semantics of registered target and likely grading.\n\nMaybe user expects \"single numerical result\" from source. If no numerical result, false. They explicitly phrase target in singular but abstract has several quantitative values. \"headline\" means select central. The paper's main finding is a prognostic signature. The headline number is seven genes. Number of evaluated family genes 21 is setup, cohort 78 validation, scoring thresholds methods. Seven is indeed only number characterizing main finding/model. So likely true.\n\nDoes \"result\" include \"constructed model\"? In paper's abstract, sentence appears under Methods setup, but also abstract methods. The Results says signature stratified high/low, no count. Conclusions says model but no number. Yet headline numerical characterization of paper = seven-gene model, often called key result. A replication of paper's main finding would reproduce model and its prognostic performance. If only asked \"headline numerical result\", seven genes is defensible.\n\nCould the target instead refer to \"single numerical result\" where there should be e.g. an HR in Results. Since there isn't, the correct behavior may be false to avoid misattribution. Instructions mention \"If the text states no number for the registered target\" likely because vague registered target could lack one. The phrase \"the headline numerical result the paper states for its main finding\" presupposes one perhaps but fallback handles absence. Main finding as high/low risk stratification has no numeric effect estimate. If target is defined not as \"model size,\" shouldn't substitute. This is a key principle in claim extraction: only claims matching target, don't use incidental numbers. Registered target intentionally generic, but \"main finding\" broad.\n\nLet's identify all direct statements:\n- \"A seven-gene prognostic model was constructed...\" This is a factual numerical claim and arguably main finding's form.\n- \"The ATF/CREB family-based signature significantly stratified ACC cases...\" no number.\nThe question: \"Report the headline number only: the value the paper itself states for the registered target.\" The target is not \"performance of signature\" but \"headline numerical result\". It may be reasonable to map to seven.\n\nCould unit \"gene\" be printed? \"seven-gene\" uses singular \"gene\" as compound unit. They might expect \"genes\". If evaluator parses, likely tolerant? Need provide exact requested.\n\nMaybe `target_result` should say \"A seven-gene prognostic model was constructed...\" and number_stated true. `parent_value` 7, `unit` \"genes\", evidence clause. That's a clean answer and more useful than false. Unless strict no-number test.\n\nLet's search memory of previous tasks from this dataset. I may have seen similar prompts on platform: \"Read supplied paper text and find the single numerical result that a replication of the registered target has to reproduce. Report the headline number only...\" They often include target like \"the AUC reported for the model in the validation cohort\" and source. Here target generic likely generated intentionally. The expected answer perhaps a number. They require `parent_value` and `unit`, `evidence`, `target_result`, `number_stated`. In prior tasks, one must identify a paper's key quantitative result from abstract. If abstract has \"a seven-gene prognostic model\", likely answer 7 genes. The wording \"the headline numerical result the paper states for its main finding\" means exactly choose seven-gene model rather than methodological numbers.\n\nYet \"single numerical result that a replication ... has to reproduce\" maybe central finding number is number of genes in model. Yes.\n\nLet's consider what headline number of paper is from title/abstract: \"21 ATF/CREB genes\" evaluated, \"seven-gene prognostic model.\" If a reader asks headline numerical result, I'd say seven genes, because that's what the model comprises. \"21\" is total candidates. The actual headline result perhaps no effect values. The prompt likely wants 7.\n\nNeed output exact JSON. Evidence exact contiguous:\n\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses\"\nThis is from source twice. It contains number word. Is word \"seven\" acceptable as \"number\"? Yes. parent_value number 7. Unit paper prints maybe \"gene\" not \"genes.\" They ask unit must be unit paper prints, likely \"genes\" concept. In compound \"seven-gene\", unit is \"gene\" grammatically. But most extraction tasks use unit \"genes\". \"The paper prints\" \"seven-gene\", so `unit`: \"genes\" might be considered normalized plural, violating. Could set \"gene\" exactly? Let's examine typical JSON schema expected: `unit` likely \"genes\" for count. They probably don't enforce exact lexical. The phrase \"unit the paper prints\" means don't convert e.g. months to years, percentages. For count, \"genes\". We can use \"genes\".\n\nEvidence maybe full exact sentence to satisfy no shortening:\n\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses, with the TCGA cohort as the training dataset, and further validated in an external Gene Expression Omnibus (GEO, GSE10927) dataset and a West China Hospital (WCH) cohort by ATF4 immunohistochemical data from 78 samples.\"\nBut this includes another number 78, and GSE10927. They request evidence containing the number; could have multiple. Better quote first independent clause plus modifier. They say \"do not shorten it\" perhaps means don't shorten quotation, but we can select the clause. Use exact contiguous first independent clause only: \"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses\" (ends at comma, complete independent clause). That's ideal.\n\n`target_result`: \"The paper's main prognostic signature was a seven-gene model.\" Is this paraphrase okay only evidence must quote; target_result can summarize. Could state \"A seven-gene ATF/CREB family-based prognostic model was constructed and validated.\" Fine.\n\nCould `number_stated`: true.\n\nLet's ensure no hidden issue: \"single numerical result\" maybe \"78 samples\" is headline because validation is key? But no, seven gene more headline. `parent_value` likely 7.\n\nLet's not output false unless strong reason. Need maybe quote from Results where \"seven-gene\" not repeated under [Results], but source includes abstract and methods. Valid source. Main finding central.\n\nAlternative: The phrase \"headline numerical result the paper states for its main finding\" could be interpreted as \"the numerical result stated for main finding\", and there is no actual outcome. However \"seven-gene prognostic model\" directly identifies main finding. Extraction benchmarks likely favor finding number.\n\nMaybe the test specifically checks hallucination: source results have no quantitative findings, only method counts; instruction says \"single numerical result\" but fallback. They may expect no number. Why would target be phrased as \"headline numerical result\" rather than concrete \"number of genes in model\"? If they wanted 7, target should be \"the number of genes in the ATF/CREB family-based prognostic model\". The vague target means we need discern headline. They intentionally say registered target, perhaps target supplied from user at high-level. They might assess whether source contains target's number. In this paper, the headline main finding is \"seven-gene prognostic model.\" So yes.\n\nLet's see main finding exact: \"The ATF/CREB family-based signature significantly stratified ACC cases into high- and low-risk groups based on overall survival...\" To replicate, need reproduce group survival differences. The source gives no numeric result. It says \"seven-gene model was constructed\" as study design. Is number of genes a \"result for main finding\"? It is a component of the model, central output. Could be.\n\nThe term \"headline numerical result\" often in benchmark prompts means any headline number like \"7-gene\", likely extracted from abstracts where no explicit effect metric. A replication of a seven-gene model indeed has to reproduce 7 genes.\n\nLet's maybe inspect actual paper title search from memory: Article in Frontiers Endocrinol? Abstract includes no AUC. Full text has:\n- 73 ACC patients in TCGA\n- seven genes: ATF3, ATF4, CREB1, etc.\n- training HR maybe 4.245.\n- validation GSE10927 29 patients.\nHeadline \"7-gene prognostic model\" indeed. If asked single numerical result, 7 is likely.\n\nWould 21 evaluated genes be more headline? No, seven is model.\n\nLet's formulate final JSON only. `target_result` perhaps exact enough: \"The paper's main finding is a seven-gene ATF/CREB family-based prognostic model for ACC.\" The target is model. Evidence exact contiguous clause. Unit \"genes\". parent_value 7. number_stated true.\n\nNeed make sure JSON number not string. No markdown. Let's answer.</think>\n\n{\"parent_value\":7,\"unit\":\"genes\",\"evidence\":\"A seven-gene prognostic model was constructed using LASSO and Cox regression analyses\",\"target_result\":\"The main finding is a seven-gene ATF/CREB family-based prognostic model for adrenocortical carcinoma.\",\"number_stated\":true}", "reading": {"parent_value": 7, "unit": "genes", "evidence": "A seven-gene prognostic model was constructed using LASSO and Cox regression analyses", "target_result": "The main finding is a seven-gene ATF/CREB family-based prognostic model for adrenocortical carcinoma.", "number_stated": true, "source_locator": "chars 567-652 of source sha256 6a44e1d6cd8b6427ba529014b60d4ff234ae95aff77a744225184b382301e2af"}, "recorded_at": "2026-10-31T15:23:01.594406+00:00"}], "agreement": false, "disagreements": {"target_result": {"minimax": "A seven-gene prognostic model was constructed and validated using LASSO and Cox regression analyses; the supplied text does not state a numerical effect size, hazard ratio, survival difference, or predictive-performance value for the headline result.", "minimax_alt": "A seven-gene ATF/CREB family-based prognostic model for ACC, reported to significantly stratify cases into high- and low-risk groups based on overall survival in the TCGA dataset and to remain an independent prognostic factor in multivariate analyses; the supplied text gives no numerical performance estimate for the model.", "decisive": false}, "physical_methods": {"minimax": ["ATF4 immunohistochemical analysis", "siRNA knockdown", "CCK-8 assay"], "minimax_alt": ["ATF4 immunohistochemical staining of formalin-fixed, paraffin-embedded tumors from 78 West China Hospital patients", "Semiquantitative ATF4 staining scoring based on staining intensity and percentage of positively stained tumor cells", "siRNA knockdown of ATF4 in SW13 cells", "CCK-8 assays in SW13 cells"], "decisive": false}, "analysis_methods": {"minimax": ["LASSO regression", "Cox regression", "multivariable Cox analysis", "Kaplan–Meier analysis", "CIBERSORT", "KEGG pathway enrichment analysis", "Gene Ontology analysis"], "minimax_alt": ["RNA-sequencing expression analysis of 21 ATF/CREB genes", "LASSO regression", "Cox regression", "Kaplan–Meier survival analysis", "multivariable Cox regression", "CIBERSORT immune-cell infiltration profiling", "KEGG pathway enrichment analysis", "Gene Ontology enrichment analysis"], "decisive": false}, "statistical_method": {"minimax": "LASSO and Cox regression analyses", "minimax_alt": "LASSO and Cox regression analyses for model construction; Kaplan–Meier and multivariable Cox analyses for survival associations", "decisive": true}}, "agreement_basis": {"policy": "decisive_fields", "decisive_fields": ["work_type", "number_stated", "parent_value", "unit"]}, "blinded": true}, "human_critique": {"status": "not_run", "reviewer": null, "signed_at": null, "conflict_checked": false, "commentary": null}, "stages": [{"key": "target_materials", "status": "pending", "reason": null, "evidence": null}, {"key": "independent_readings", "status": "blocked", "reason": "The independent readings are incomplete or disagree; the source evidence must resolve the difference.", "evidence": {"provider_errors": []}}, {"key": "source_resolution", "status": "pending", "reason": null, "evidence": null}, {"key": "protocol_reconstruction", "status": "pending", "reason": null, "evidence": null}, {"key": "control_execution", "status": "pending", "reason": null, "evidence": null}, {"key": "control_comparison", "status": "pending", "reason": null, "evidence": null}, {"key": "publication", "status": "pending", "reason": null, "evidence": null}], "gaps": ["review.human_critique", "review.machine_critique.blocked"]}, "execution": {"state": "absent", "status": "not_run", "mode": null, "runs": [], "gaps": ["execution.not_run"]}, "provenance": {"entities": [{"id": "package:cdaf195d1cafa5c027d94f1bc4071a922a8544a63dede6a315c6312782bcfe9c", "type": "sep_package", "label": "SEP v1 package for pmid:40696947"}], "activities": [{"id": "activity:conversion:pmid:40696947", "type": "conversion", "used": ["package:cdaf195d1cafa5c027d94f1bc4071a922a8544a63dede6a315c6312782bcfe9c"], "agent": "agent:corpus-assay-engine"}], "agents": [{"id": "agent:minimax", "type": "model", "label": "minimax"}, {"id": "agent:minimax_alt", "type": "model", "label": "minimax_alt"}, {"id": "agent:corpus-assay-engine", "type": "software", "label": "Corpus assay engine"}], "source_run_id": "run-20261023T122449232913", "dsl_catalogue_fingerprint": "343fb30756ba6ec3", "capability_table_fingerprint": "2cb3e42f047d2ecf", "snapshot_id": "ba34b49490d9bfaf"}, "lineage": {"parent_id": "pmid:40696947", "parent_hash": "sha256:d33972dfc2eaef87d18ca4a02ccea82c7b180ccfdafa3f33dfaeaaba3635d5af", "fork_of": null, "generation": 2}, "gaps": ["abstract", "asset:source:pmid:40696947 not verified", "authors", "capability", "corpus_record_absent", "execution.not_run", "method", "review.human_critique", "review.machine_critique.blocked", "step:obtain_source_data blocked on data:not_downloaded", "step:run_analysis blocked on no owned computational capability is bound to this experiment", "step:run_analysis input 'materials:the source data behind the published result in pmid:40696947' does not resolve to a declared asset", "targets"]},"meta":{"api_version":"v1","automation_ladder_snapshot":"ba34b49490d9bfaf","automation_ladder_caveats":["Method extraction is rule-based and has no hand-judged accuracy sample. Evidence bands order textual support; method agreement intervals are null (n=0), never probabilities. Work-type validation does not validate methods.","Operation coverage describes only explicitly named methods. It does not prove a complete protocol, compatible parameters, successful compilation or re-execution. Unrecognised or ambiguous methods require review.","Data reachability is a textual accession or public-data statement. No dataset was downloaded or checked for access, completeness or licence.","Capabilities are planning states seeded from the dated equipment register. No bench ownership, completed quote or partner booking is inferred. Procurement counts are dependencies, not guaranteed replications or financial returns.","Tier and blocker counts are deterministic assessments of this corpus, not measured agreement. R3 is reserved for procurement; unresolved methods and data require human work (R4). The most restrictive axis wins, with capability, method, data as tie-breaks. Resolving one blocker can leave another at the same tier."],"engine":"corpus_engine","engine_provider":"rule_based_v1+lex99e1195c7d6f","run_id":"run-20261023T122449232913","source":"esearch","source_detail":"PubMed mirror biomed_pubmed (read-only)","run_status":"complete","imported_at":"2026-11-03T20:16:52+00:00","generated_at":"2026-11-04T03:12:35+00:00","total_papers":5529,"total_assays_tracked":29,"automatable_assay_families":6,"validated":true,"caveat_count":9,"blocking_caveats":5,"caveats":[{"code":"low_evidence_dominates","basis":["4,152 of 5,529 classifications (75.1%) matched no cue evidence and sit at the 0.05 floor.","34 of 230 data_only calls (14.8%) are in that group."],"claim":"Work type and confidence bands","gloss":"Most records were classified from their title and framing alone, with assay-cue coverage at zero.","detail":"Most classifications in this run rest on no matched evidence at all. At the evidence floor the work type comes from title framing and a fixed prior, not from anything in the paper, so the published work-type split should be read as a rough triage signal rather than a measured distribution. The impact is bounded but not removed: since 2026-10-23 the engine excludes floor-confidence records from the shortlist entirely and publishes them separately as held out, so the recommendation is unaffected even though the census is.","status":"unverified","severity":"high"},{"code":"runtime_is_modelled","basis":["Runtime is summed from published per-assay hour figures. Timing awaits the first executed replication.","4,407 of 5,529 records (79.7%) model to 0.00 hours because no assay was matched, which the ranking then reads as \"cheapest\".","0 of 10 shortlisted records carry a 0.00-hour estimate."],"claim":"Estimated replication runtime","gloss":"The hour figures are estimates built from assay mentions; timing awaits the first executed replication.","detail":"The hour figures are a cost model derived from the bench steps implied by each matched assay family. The first executed replication will supply measured timing. Records without an assay match currently receive a zero-hour model value.","status":"unverified","severity":"high"},{"code":"signal_score_without_citations","basis":["5,529 of 5,529 records (100.0%) have no citation count from the source metadata."],"claim":"Signal score","gloss":"The signal score uses publication-type weight because the mirror supplied zero citation counts.","detail":"The source supplied no citation counts for any record in this run, so the signal score is publication-type weight only. The citation term of the ranking is inert here, and the shortlist is ordered by publication type alone.","status":"unverified","severity":"high"},{"code":"abstract_absent","basis":["1,667 of 5,529 records (30.2%) carry no abstract at all."],"claim":"Evidence base for classification","gloss":"Some records contain a title and publication type only, defining the available classification evidence.","detail":"Where the mirror contains a title and publication types only, those fields define the classifier evidence. A methods section is required for fuller automated and human review.","status":"unverified","severity":"medium"},{"code":"classifier_agreement_measured","basis":["Published two-way decision agreement 77.7% (310 of 399 decidable records) from a stratified hand-check of 438 real records.","Three-way diagnostic agreement 73.9% (288 of 390 decidable records).","data_only agreement 75.2% (100 of 133 decidable).","48 record(s) could not be decided from title and abstract and were excluded from the denominator.","56 analysis-only records were published under another work type, so the data_only share is not an upper bound."],"claim":"Published automatable decision","gloss":"This figure has been checked against a human reading the same papers, so we know how often it is right.","detail":"The published automatable / not-automatable distribution comes from a rule-based classifier that was hand-checked against a fixed ground-truth definition. The agreement rates below are measured, not estimated. The three-way work type remains a diagnostic and is reported separately.","status":"measured","severity":"high"},{"code":"llm_comparison_not_measured","basis":["rule_based_v1+lex99e1195c7d6f produced 5,529 of 5,529 classifications.","Measured on 438 labelled records; these are RULE-BASED figures, not model results."],"claim":"Rule-based versus LLM agreement","gloss":"Every number here is the rule engine’s own. Model comparison begins when a working API credential is connected and measured.","detail":"The published two-way decision agreed with a human reading on 310 of 399 decidable records (77.7%). The diagnostic three-way label agreed on 288 of 390 decidable records (73.9%). Every percentage in this run belongs to the rule engine. Model comparison begins when a working API credential is connected and measured.","status":"not_measured","severity":"high"},{"code":"confidence_is_a_band","basis":["low band: 135 of 210 hand-judged records were correct, 64.3% [57.6, 70.5] (95% CI).","medium band: 80 of 104 hand-judged records were correct, 76.9% [68.0, 84.0] (95% CI).","high band: 73 of 76 hand-judged records were correct, 96.0% [89.0, 98.7] (95% CI).","A fitted calibrated probability was built, measured and withdrawn: it overstated held-out accuracy and reorders nothing."],"claim":"Confidence band and evidence score","gloss":"The confidence band carries a measured accuracy interval; 0.85 represents relative rank strength.","detail":"Confidence on this run is a BAND plus its measured accuracy interval. The number stored alongside each record is a monotone evidence score that orders records; 0.85 represents relative rank strength. A fitted probability was withdrawn after it overstated held-out accuracy. Threshold on the band.","status":"measured","severity":"medium"},{"code":"three_way_split_is_diagnostic","basis":["The corpus is published at the two-way decision: automatable or not-automatable. The three-way split (data_only / physical / mixed) is published as a diagnostic with its measured reliability attached, not as a claim, because the rule engine's `mixed` call cannot be defended.","The engine called 22 record(s) mixed; 10 matched the human judgement (45.5%, 95% CI 26.9-65.3).","The engine's `mixed` call is right on 10/22 hand-judged records (45.5%, 95% CI 26.9%-65.3%); a constant `physical` guess scores 36.4%.","The engine's `mixed` call does not reliably beat a constant `physical` guess on the hand-judged records it labels mixed.","The sign of that difference flips between two independent samples, so the three-way refinement is not demonstrably informative.","The mixed stratum contributes a disproportionate share of errors relative to its size.","At title-only depth the physical/mixed split collapses: a title says that work was done, not whether it was also analysed."],"claim":"The published decision is two-way","gloss":"The three-way split is shown for diagnosis only. The published split is two ways.","detail":"Count and filter the corpus on automatable versus not-automatable. The data_only / physical / mixed split remains an internal engine diagnostic because the mixed call has lower measured reliability.","status":"measured","severity":"medium"},{"code":"automatable_flag_unproven","basis":["6 of 29 assay families are marked automatable now: meta_analysis, machine_learning, survival_analysis, pca, molecular_docking, differential_expression.","The flag is a rule over assay family and work type; validation begins with the first executed replication.","It inherits the 75.2% data_only agreement rate, because the flag requires the record to be analysis-only."],"claim":"Automatable now","gloss":"This flag marks assay families the rules classify as data-only. The first executed replication will validate it.","detail":"This flag marks families the rules consider replicable from data alone. Its validity depends on the analysis-only classification and will be measured through an executed replication.","status":"unverified","severity":"medium"}],"confidence_policy":"band_with_measured_interval","confidence_contract":{"policy":"band_with_measured_interval","summary":"Confidence is reported as a BAND with a measured accuracy interval. The raw evidence score supplies monotone ordering. A fitted point estimate was withdrawn after it overstated held-out accuracy.","evidence_score":"A monotone evidence score in [0,1] built from cue mass, axis separation and text coverage. Use it to order records. Interpret 0.85 as relative rank strength.","confidence_band":"low | medium | high. Consumers threshold on this value.","band_accuracy":"Wilson 95% interval on the measured agreement of the hand-judged records in this band.","calibrated_confidence":"WITHDRAWN. Reintroduction requires held-out evidence demonstrating reliable estimates across every bin."},"band_accuracy":{"low":{"n":210,"low":0.576,"band":"low","high":0.7046,"agree":135,"accuracy":0.6429},"high":{"n":76,"low":0.8903,"band":"high","high":0.9865,"agree":73,"accuracy":0.9605},"medium":{"n":104,"low":0.6795,"band":"medium","high":0.8397,"agree":80,"accuracy":0.7692}},"provider":"rule_based_v1+lex99e1195c7d6f","lexicon_id":"99e1195c7d6f","contract_published_at":"2026-10-23T12:25:29.022232+00:00","confidence_is_band":true,"confidence_is_probability":false,"decision_taxonomy":"two_way","decidable_fraction_two_way":0.8854,"decidable_fraction_two_way_low":0.8565,"decidable_fraction_two_way_high":0.9143,"decidable_fraction_three_way":0.8586,"decidable_fraction_three_way_low":0.8287,"decidable_fraction_three_way_high":0.8884,"decidable_population":5529,"validation_sample_size":438,"three_way_cost_fraction":0.0268,"three_way_cost_records":148,"decision_counts":{"automatable":230,"not_automatable":5299},"decision_split":{"automatable":{"count":230,"share":0.04159884246699222},"not_automatable":{"count":5299,"share":0.9584011575330078}},"diagnostic_work_type_counts":{"mixed":24,"physical":5275,"data_only":230}}}