Corpus 1,286 assessed · 1,187 scored · 648 reproduced ≥75 · 174 flagged ·∅ 73.9/100
← New search

Research and experimental verification on the mechanisms of cellular senescence in triple-negative breast cancer.

PeerJ · 2024
L1 67/100 PQI 89
Why this verdict

The main results reproduced, with only marginal, non-material deviations.

Reproduced on the brainbox compute brainarbeit.com
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q5 · Derivability / plausibility 🟡
Content-critical question only partially held
+2 pts
From: Q7 · Core claim 🟡
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q4 · Cause of the deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q2 · Endpoint comparability 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +7
✓ What held up
  • Any deviation was negligible
What did not (or only partly)
  • 🟡Could not use the authors’ exact input data
  • 🟡Reported values were only indirectly comparable
  • 🟡A deviation arose in the data or preprocessing
  • 🟡A deviation was attributed to the published material
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
67/100
Reproducibility score
0.4 SD below mean
vs. all fields · 1187 studies
🎯 Scores higher than 30% of all assessed papers rank 801 of 1187 scored

A 0–100 reproducibility-quality score from the per-question grades, shown as a z-score: standard deviations above (+) or below (−) the mean of comparable assessments.

Reproduction agent’s raw note

Described well enough at the cohort-construction level; partial 1:1. Reproduced THREE clearly-specified, pipeline-derived data points from shipped + public data, each matching the authors' own in-script comments: (C1) the 253-gene senescence set is the union of the shipped gmt's KEGG_P53_SIGNALING_PATHWAY and REACTOME_CELLULAR_SENESCENCE pathways — exact; (C2) the TCGA TNBC cohort — 116 ER/PR/HER2-negative patients → 114 with OS.time>0 → 113 after intersecting expression, reproduced independently from the shipped clinical file; (C3) GSE58812 = 107 samples, confirmed exactly from public GEO. NOT attempted (the hard ~20%): the 186-gene TCGA intersection, 69 univariate-Cox genes, the 4-gene LASSO prognostic model (MMP28/CT83/ACP5/KRT6A) and its ROC AUCs, the 3 consensus-clustering subtypes, CNV/GISTIC, immune deconvolution, drug sensitivity, nomogram — all chain through an UNSHIPPED Sangerbox TCGA TPM matrix (Merge_TCGA-BRCA_TPM.txt, 25,483-gene universe) plus a private Java GSEA tool and private /pub1/... code paths. An independent TCGAbiolinks pull would use a different gene universe and would not be a 1:1 reproduction, so per the 80/20 brief it was documented rather than chased. No value was found to be non-derivable/fabricated; the reproduced cohort numbers support the upstream figures as genuine. Wet-lab validation (qRT-PCR, siRNA, transwell) is out of scope. NOT a drop: the paper is described well enough to reproduce its data-assembly steps, which check out.

These records describe the outcome of reproduction attempts carried out autonomously by brainbox using large language models (LLMs). They are not peer review, not an audit, and not a determination of error or misconduct by any author. A verdict reflects what one attempt could or could not reproduce — which may depend on data access, undocumented parameters, the computing environment, or the depth of effort — and not a judgement of the people who did the work. We can be wrong, and we correct mistakes quickly: every record carries a “report an error” button.

Assessment versions

Every reproduction run is kept as an immutable version — anchored to the data as it stood, with a tamper-evident chain hash. A rerun (e.g. after an author updates a deposit) adds a new version; the previous one stays on record.

  1. v1 current initial assessment Score 67
    assessed: 2026-06-15 ⛓ 6094074c84d5
✎ I am an author of this paper

Updated or fixed a deposit, or is there an erratum? Ask us to re-run the metrics. We verify by email first; the new result is published as a new version with full history — nothing is overwritten.

Reason for the rerun

We email you a confirmation link first. The rerun is an objective re-measurement — it cannot change the verdict in your favour, only ask us to look again.

Provenance — full disclosure

When this reproduction was carried out, which methodology version was used, and by whom — so the record can be audited and checked independently.

Reproduced
2026-06-15
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-15
no human curator yet
Last updated
2026-09-19

Provisional, curator- or AI-assessed, and independently checkable. A reproduction outcome states what one attempt could reproduce — not a judgement of the authors.

Deep full-text extraction

Model: sonnet
Founding hypothesis

The authors hypothesize that markers associated with senescence features and prognosis in triple-negative breast cancer (TNBC) could influence cancer progression and prognosis through effects on tumor microenvironment homeostasis, cytokine release, and genomic mutations.

Core claims
  • TNBC can be classified into three molecular subtypes (clusters 1, 2, 3) based on cellular senescence-related pathway scores finding
  • Cluster 1 has the best prognosis followed by cluster 2 then cluster 3; gene expression levels are lowest in cluster 2 and highest in cluster 3 finding
  • Clusters 1 and 3 show a high degree of immune infiltration finding
  • TIDE scores are higher in cluster 3, indicating greater likelihood of immune escape and reduced immunotherapy benefit finding
  • A senescence-related risk model was constructed; prognostic risk genes MMP28, ACP5, and KRT6A are up-regulated while protective gene CT83 is down-regulated in TNBC cell lines, validating bioinformatic predictions finding
  • ACP5 promotes migration and invasion abilities in two TNBC cell lines finding
  • Eleven immune cell subpopulations were annotated in TNBC scRNA-seq data based on classical marker genes method
  • The constructed prognostic risk model is valid for assessing tumor microenvironment characteristics and TNBC chemotherapy response finding
Experimental setups
Assay System Perturbation Readout Platform
scRNA-seq clustering/annotation (Seurat, TSNE) TNBC tumor tissue, 9 samples (GSE176078) none cell subpopulation identity and senescence pathway scores Seurat package
ssGSEA senescence pathway scoring TCGA TNBC and GSE58812 bulk RNA-seq tumor/para-cancerous tissue none cellular senescence-related pathway enrichment scores GSVA package
Unsupervised consensus clustering TCGA TNBC bulk RNA-seq (113 tumor samples) none molecular subtype assignment (clusters 1-3) ConsensusClusterPlus
Immune infiltration and immunotherapy response scoring TCGA TNBC bulk RNA-seq none immune/stromal scores, immune cell composition, TIDE immunotherapy score ESTIMATE, MCP-counter, TIMER, EPIC, TIDE
Mutation analysis (CNV and SNV) TCGA-TNBC dataset none copy number variation and single nucleotide variant profiles gistic2, mutect2, maftools
Univariate Cox and LASSO regression TCGA TNBC bulk RNA-seq none prognosis-related feature genes and risk score glmnet package
qRT-PCR MDA-MB-468, MDA-MB-231 (TNBC) and MCF10A (non-tumorigenic) cell lines none mRNA expression of MMP28, CT83, ACP5, KRT6A LightCycler 480 PCR System
Transwell migration/invasion assay MDA-MB-468 and MDA-MB-231 TNBC cell lines ACP5 siRNA knockdown vs negative control number of migrated/invaded cells
Key results
  • TNBC samples classified into three senescence-based molecular subtypes (clusters 1, 2, 3)
  • Cluster 1 showed best prognosis, cluster 3 worst; gene expression lowest in cluster 2, highest in cluster 3
  • Clusters 1 and 3 exhibited high immune cell infiltration
  • TIDE scores were higher in cluster 3 patients, indicating greater immune escape and lower immunotherapy benefit
  • MMP28, ACP5, and KRT6A were up-regulated and CT83 down-regulated in TNBC cell lines by qRT-PCR
  • ACP5 knockdown/expression altered migration and invasion in MDA-MB-468 and MDA-MB-231 cells, indicating ACP5 promotes these abilities
  • 38,007 of 38,582 filtered cells passed QC and were clustered into 11 subpopulations
  • Risk model effectively assessed TME characteristics and chemotherapy response in TNBC
Key statistics
  • count 38,582 cells filtered; 38,007 cells retained after QC (scRNA-seq cell filtering across 9 samples (GSE176078))
  • count 113 para-cancerous and 113 tumor samples (TCGA TNBC clinical/expression dataset)
  • count 107 qualified tumor samples and 16,416 genes (GSE58812 microarray dataset after filtering)
  • other clustering resolution = 0.1 yielding 11 subpopulations (scRNA-seq FindClusters parameters)
  • other logfc = 0.5, Minpct = 0.35, adjusted P < 0.05 (thresholds for marker gene screening (FindAllMarkers))
  • pvalue p < 0.05 (significance threshold for univariate Cox prognosis-related gene screening)
  • other 500 bootstraps, 80% of patients per bootstrap (consensus clustering parameters (ConsensusClusterPlus))
  • count 3 independent experiments per group (replication for wet-lab experiments (qRT-PCR, transwell))

Statistical methods review

Model: sonnet

A neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.

This bioinformatics-plus-wet-lab study integrated scRNA-seq (GSE176078, 9 samples, 38,007 cells after QC) and bulk RNA-seq data (TCGA TNBC n=113 tumor; GSE58812 n=107) to classify TNBC into three senescence-based molecular subtypes via ssGSEA pathway scoring and unsupervised consensus clustering. Prognostic gene selection used univariate Cox regression followed by LASSO-Cox with 10-fold cross-validation to build a risk score, evaluated by Kaplan–Meier/log-rank and time-dependent ROC. In vitro hub-gene validation (n=3 independent experiments) used qRT-PCR and transwell assays, with two-group differences assessed by Wilcoxon and three-group differences by Kruskal–Wallis.

Replicationmixed Sample sizeStatistical power reported post-hoc as 1.0 (RNASeqPower) for both GSE176078 and TCGA datasets; wet-lab experiments stated as n=3 independent experiments per group; no a priori power or sample-size justification described GroupsThree senescence-based TNBC clusters (1, 2, 3); high-risk vs low-risk by Riskscore; tumor vs para-cancerous tissue; TNBC cell lines (MDA-MB-468, MDA-MB-231) vs normal epithelial line (MCF10A); ACP5 siRNA vs negative control Pairingunclear Randomization/blindingnot stated Dispersionunclear Effect sizesno Multiplicity correctionAdjusted p-value threshold (adjusted p < 0.05) stated for scRNA-seq marker gene screening via FindAllMarkers; no explicit correction described for the many parallel Wilcoxon pathway tests or the univariate Cox gene-screening step
Statistical tests used
Test Applied to n Assumptions
ssGSEA (single-sample gene set enrichment analysis, GSVA package) Senescence-related pathway scoring per cell (scRNA-seq) and per bulk sample (TCGA, GSE58812); also for G1/S, G2M checkpoint, telomere extension, and EMT pathway scores in TCGA 38,007 cells (scRNA-seq); 113 tumor + 113 para-cancerous (TCGA); 107 tumor samples (GSE58812) not stated
Wilcoxon rank-sum test (wilcox.test) Significance of each senescence-related pathway in cancer vs para-cancerous tissues (bulk RNA-seq); general two-group comparisons 113 tumor vs 113 para-cancerous (TCGA); 107 tumor samples (GSE58812) not stated
Kruskal–Wallis test (kruskal.test) Immune cell infiltration score differences across three TNBC clusters (MCP-counter, TIMER, EPIC outputs) 113 TNBC samples (TCGA) not stated
Univariate Cox proportional-hazards regression (coxph, survival package) Screening prognosis-related genes at p < 0.05 for LASSO input 113 TNBC samples (TCGA) not stated
LASSO-Cox regression with 10-fold cross-validation (glmnet package) Feature selection and construction of the prognostic Riskscore model 113 TNBC samples (TCGA) not stated
Log-rank test with Kaplan–Meier curves Overall survival comparison between high-risk and low-risk groups; prognosis across the three clusters 113 TNBC samples (TCGA); 107 samples (GSE58812) not stated
Pearson correlation (rcorr, Hmisc package) Correlation between risk score and immune cell infiltration scores 113 TNBC samples (TCGA) not stated
limma moderated t-statistic (linear model) Differential expression analysis across clusters 1, 2, and 3 in TCGA and GSE58812 (|log2FC|>1, p<0.05) 113 TNBC samples (TCGA); 107 samples (GSE58812) not stated
Unsupervised consensus clustering (ConsensusClusterPlus; hc algorithm, canberra distance, 500 bootstraps, 80% resampling per bootstrap) Identification of optimal TNBC molecular subtype number (k tested 2–10, k=3 chosen via consensus matrix and CDF) 113 TNBC samples (TCGA) not stated
Approaches that could also have been used
  • ssGSEA was applied to score senescence-related pathways in individual cells from scRNA-seq data
    Could also: AUCell or UCell could also score pathway activity at single-cell resolution — AUCell and UCell are designed for the sparse, zero-inflated distribution characteristic of scRNA-seq and rank cells by the relative expression of a gene set within each cell's own detected transcriptome, which can be more stable than ssGSEA when sequencing depth varies widely across cells
  • Multiple senescence-related pathways were each tested individually with a Wilcoxon test comparing cancer vs para-cancerous tissue
    Could also: Benjamini–Hochberg FDR correction (or Bonferroni) applied across all pathway tests could also be used — When many pathways are tested simultaneously, controlling the false-discovery rate limits the expected proportion of spurious findings; reporting adjusted q-values alongside raw p-values is standard practice in multi-pathway analyses
  • Kruskal–Wallis was used to compare immune cell scores across three clusters without an explicit pairwise post-hoc procedure described
    Could also: Dunn's test or pairwise Wilcoxon with Benjamini–Hochberg correction could also follow a significant Kruskal–Wallis result — Post-hoc pairwise comparisons identify which specific cluster contrasts drive the overall significance, adding interpretive precision without inflating the family-wise Type I error rate
  • Pearson correlation was used to relate risk scores to immune cell infiltration scores
    Could also: Spearman rank correlation could also be used for this association — Spearman correlation is robust to non-normal distributions and outliers, which are common in ssGSEA enrichment scores and immune deconvolution estimates; both are standard choices, and stating which was selected (and checking the distributional assumption) aids reproducibility
  • limma (developed for microarray and normalized expression data) was applied for differential expression in both the microarray dataset (GSE58812) and the RNA-seq dataset (TCGA)
    Could also: DESeq2 or edgeR with negative-binomial modeling, or limma-voom, could also be applied to the raw-count RNA-seq data in TCGA — DESeq2 and edgeR explicitly model count overdispersion; limma-voom adapts limma's framework to count data via precision weights — these approaches are widely recommended for RNA-seq and may yield a different DEG list, particularly at low counts
  • Wet-lab validation (qRT-PCR, transwell) used n=3 independent experiments; the statistical test applied to compare groups and the dispersion measure used for figures are not described in the statistical analysis section
    Could also: Student's t-test or Mann–Whitney U with explicitly reported mean ± SD (or median ± IQR) could also be stated for these comparisons — Naming the exact test, reporting a measure of spread alongside the central tendency, and providing exact p-values for each wet-lab comparison allows readers to independently assess the precision and reproducibility of the in vitro findings
Software: R 4.2.0 · Seurat · GSVA (ssGSEA) · ConsensusClusterPlus · glmnet (LASSO) · survival (coxph) · timeROC · limma · maftools · MCPcounter · Hmisc (rcorr) · copycat (CNV inference from scRNA-seq) · RNASeqPower · gistic2 · ESTIMATE · TIDE · TIMER · EPIC

Citation network

Where this publication sits in the reproducibility-weighted citation graph — what it is built on, and what is built on it. Citation data from OpenAlex.

Citations
2
Impact: low
Foundation confidence
None of its references are in our reproducibility record yet — its foundation cannot be assessed.
Topics

No assessed neighbours yet — the network grows as more papers are assessed.

Data lineage

The datasets this paper uses (text-mined from the full text via Europe PMC), and which other assessed papers stand on the same data. A shared dataset is a factual link — not a judgement.

GSE176078 GEO in Data Availability (http://purl.obolibrary.org/obo/IAO_0000611)
also used by 3 papers:
GSE58812 GEO in Data Availability (http://purl.obolibrary.org/obo/IAO_0000611)
no other assessed paper uses this yet

Downstream reach in the literature

98 downstream papers · 1 datasets

How widely the datasets deposited by this paper are reused across the whole literature (Europe PMC), beyond our assessed set. This is a factual dependency map — reusing a public dataset is normal, good science. It is not a judgement on the downstream papers; the only verdict here is this paper's own, with its cited rationale.

What was reproduced

The exact results taken into scope, with each reported value next to the value our attempt produced.

Scope — pmid-38435998

Paper: Cao T, Huang M, Huang X, Tang T. Research and experimental verification on the mechanisms of cellular senescence in triple-negative breast cancer. PeerJ 2024. PMID 38435998 · PMCID PMC10909353 · DOI 10.7717/peerj.16935

Code: https://github.com/ctf1985/Raw-and-Experimental-Data commit 61ecd5957ca1adc209362f36d3026dee96b8b602 (only commit; pushed 2023-11-17; no license). Data: GEO GSE176078 (scRNA), GSE58812 (bulk validation), TCGA-BRCA, METABRIC.

What the repo ships

  • One monolithic R pipeline: scripts/20220518_TNBC.cellAge.scRNA.R (3081 lines).
  • origin_datas/cellAge.pathway.gmt — 17 senescence/aging pathways (GO_BP, Reactome, KEGG), 539 unique genes.
  • origin_datas/TCGA/…clinical.txt (TCGA-BRCA clinical, 1097 patients).
  • origin_datas/METABRIC/… (METABRIC clinical + agilent microarray).
  • The authors' intermediate result files (tcga.subtype.txt, tcga.group.txt, tcga.cellage.score.txt, …) and per-figure PDFs.
  • The raw data of experiments/ — wet-lab qRT-PCR / transwell raw data (OUT OF SCOPE: manual/wet-lab).

What the repo does NOT ship (blocks 1:1 of the deep claims)

  • origin_datas/TCGA/Merge_TCGA-BRCA_TPM.txt — the TCGA expression matrix the whole prognostic analysis reads (script line 678). Absent. Authors built it on the Sangerbox platform; it has a non-standard gene universe (25483 genes).
  • Merge_GeneLevelCopyNumber.txt (CNV) — absent.
  • Private absolute paths to the authors' server: source('/pub1/data/mg_projects/projects/codes/mg_base.R'), /pub1/data/mg_projects/TCGA/Matrix/cnvs/…, a custom Java GSEA jar (MG_GSEA.jar), and Sangerbox helper functions (getGEOExpData, mg_RunGSEA_wtl, parseGSEAResult). None are obtainable → the GSEA-derived pathway selection and several downstream steps cannot be re-run as-is.

In-scope, pipeline-derived results (attempted)

# Result Pipeline Reproducible from Decision
C1 253 senescence genes from 3 pathways union of GSEA-significant gmt pathways shipped gmt DONE
C2 TCGA TNBC cohort = 113 tumor (+113 normal) ER/PR/HER2-neg clinical filter ∩ TPM shipped clinical (+ TCGA barcode facts) DONE
C3 GSE58812 validation = 107 samples (×16,416 genes) GEO GPL570 download + probe→gene collapse public GEO DONE (107 exact; gene count = annotation-dependent)

In-scope but NOT attempted (the hard ~20%, gated on private infrastructure)

  • 186 senescence genes present in TCGA; 69 univariate-Cox genes (p<0.05); 4-gene LASSO risk model (MMP28, CT83, ACP5, KRT6A) at lambda=0.0782; ROC AUCs (0.91/0.93/0.74/0.83/0.84); 3 consensus-clustering molecular subtypes; CNV/GISTIC; immune deconvolution; drug-sensitivity; nomogram.
  • Why skipped: every one of these reads the unshipped Sangerbox Merge_TCGA-BRCA_TPM.txt (and/or the private GSEA tool / CNV matrices). An independent TCGAbiolinks pull has a different gene universe (~60k Ensembl genes vs their 25,483), so the gene-count-dependent figures (186, 69, even the 4-gene LASSO selection) would not be a 1:1 reproduction — they would be a different analysis. Per brief 80/20, not chased; documented instead.

Out of scope (wet-lab / manual)

qRT-PCR (MDA-MB-468/231, MCF10A), siRNA ACP5 knockdown, transwell invasion/migration assays.

C1
Reported
253 senescence genes from 3 pathways
Reproduced
253 (KEGG_P53_SIGNALING_PATHWAY ∪ REACTOME_CELLULAR_SENESCENCE; Reactome sub-pathways subsumed)
exact
C2
Reported
TCGA TNBC = 113 tumor (+113 normal)
Reproduced
116 triple-negative → 114 with OS.time>0 → 113 after TPM intersection
within tolerance
C3
Reported
GSE58812 = 107 samples × 16,416 genes
Reproduced
107 samples (exact), GPL570, 54,675 probes; protein-coding gene-collapse not recomputed
partial
C4
Reported
4-gene LASSO model MMP28/CT83/ACP5/KRT6A (λ=0.0782); AUC 0.91/0.93/0.74/0.83/0.84
Reproduced
not attempted
partial
C5
Reported
186 senescence genes in TCGA; 69 univariate-Cox genes (p<0.05)
Reproduced
not attempted
partial

Assessments & scoring basis

Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.

🤖 AI curator · claude (ai-curator room) · v1.0 L1 67/100

An automated assessment. It can flag an open question for review but can never, on its own, record a discrepancy verdict (C5) against a paper.

🟡1. Data identity
🟡2. Endpoint comparability
🟡3. Location of the main deviation
🟡4. Cause of the deviation
🟡5. Derivability / plausibility
🟢6. Severity of the deviation
🟡7. Core claim
🟡8. Severity of the miss (overall human judgment)
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q5 · Derivability / plausibility 🟡
Content-critical question only partially held
+2 pts
From: Q7 · Core claim 🟡
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q4 · Cause of the deviation 🟡
Minor / cosmetic deviation
+1 pts
From: Q2 · Endpoint comparability 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +7

Three clearly-specified, pipeline-derived figures reproduce cleanly: the 253-gene senescence set is the exact union of KEGG_P53_SIGNALING_PATHWAY ∪ REACTOME_CELLULAR_SENESCENCE, the TCGA TNBC cohort rebuilds to 113 (116→114→113) from shipped clinical data, and GSE58812 confirms 107 samples exactly. The gaps are on data-availability and our-scope sides, not a demonstrated authors' error: the entire prognostic chain (186/69 Cox genes, the 4-gene LASSO model and its AUCs) reads an unshipped Sangerbox TPM matrix plus a private GSEA tool, so those values cannot be put 1:1 against our output and the central prognostic claim stays untested. No value was flagged as non-derivable-in-principle or 'too perfect' — severity of what was checked is negligible, but coverage of the core conclusion is limited.

🤝
Reproduced automatically — and fairly

Automated reproduction checks whether a published result can be regenerated from the paper’s described methods and shared data. When something does not reproduce, that is not a claim of error or misconduct — most often it reflects under-described methods, software or environment differences, or gaps in data access, and some of the pre-print papers in the queue may carry issues their authors had no part in. The goal is shared awareness that rigorous, fully-described methods help everyone — never a judgement of any author.

Are you an author? We would genuinely like to hear from you — to clarify the record, add data or code, re-run the pipeline after an accession update, and publish your response right next to the assessment. Everything here is open and auditable.

🚩 Report an error in this record

Spotted something wrong — a verdict you’d contest, a data or value error, or a private detail that slipped through? Tell us, with a short justification. Authors and readers are equally welcome to write in; we review every report.

Prefer email, or the form below not working? Contact us at [email protected].

Reproduction footprint

claude-opus-4-8

Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.

133.2 k
tokens (I/O) · 8.8 M incl. cache
13 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.