Corpus 1,272 assessed · 1,173 scored · 643 reproduced ≥75 · 168 flagged ·∅ 74.1/100
← New search

Unveiling prognostics biomarkers of tyrosine metabolism reprogramming in liver cancer by cross-platform gene expression analyses.

PLoS One · 2020
L1 51/100 3/4
⚑ Flagged for review — a reproduced result did not match the reported value

Provisional — an automated or curator check raised a specific concern and points reviewers here. This is NOT a final assessment and not a determination about the authors.

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: Q6 · Severity of the deviation 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +5
✓ What held up
  • Same input data as the authors
  • Reported values were directly comparable
What did not (or only partly)
  • 🔴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 deviation was non-trivial in magnitude
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
51/100
Reproducibility score
1.3 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 13% of all assessed papers rank 1018 of 1173 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 for a clean 1:1 on the pipeline output. The cited repo (nguyenquyha/IHC-method, commit 41db5bb) is the authors' own self-contained IHC quantification tool (P16) that ships input images (112 PNGs: FAH/GSTZ1/HGD/HPD), the MATLAB code (brown_calc.m), the intermediate masks, and the expected output (brown_ratio_table.csv). Ported brown_calc.m to Python and re-ran on «our HPC» (SLURM «job»). The shipped 112-value brown-ratio table reproduces essentially EXACTLY: Pearson r=0.99999971, all 112 within 1e-2, 110/112 within 1e-3, and the regenerated tissue masks overlap the shipped masks at Jaccard 0.998 — so both the color step and the morphology port are faithful. Of the three Figure-4 derived fold-changes, GSTZ1 reproduces EXACTLY (2.27-fold, p=0.0007 -> 2.269, p=0.00068) using mean_N/mean_T + unpaired t; but HGD (1.67) and HPD (2.26) are NOT recoverable from the shipped table under that same recipe (got 2.05 and 1.98) and no tested alternative (median, outlier exclusion, per-sample fold) recovers them — flagged as a possible discrepancy for human adjudication (likely a different image subset/version behind those two panels; not asserted as fabrication). NOT attempted: the paper's main cross-platform GSE89377 microarray DEG / tyrosine-metabolism / prognostic-survival analysis — no analysis code is shipped for any of it (the repo is only the IHC tool), so that ~20% is out of scope (no_code sub-result).

💻 Code ↗ 🗄 Data: GSE89377

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 51
    assessed: 2026-06-15 ⛓ da501b6985b1
✎ 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
2026-08-05

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: opus
Founding hypothesis

The study tests whether the five tyrosine catabolic enzymes (TAT, HPD, HGD, GSTZ1, FAH) are dysregulated in hepatocellular carcinoma and whether their expression has clinical and prognostic significance, hypothesizing that tyrosine metabolism is reprogrammed during HCC development.

Core claims
  • Five tyrosine catabolic enzymes (TAT, HPD, HGD, GSTZ1, FAH) are downregulated in HCC versus normal liver at mRNA and protein levels finding
  • Low expression of tyrosine catabolic genes (notably TAT, HGD, GSTZ1) correlates with poorer overall and disease-free survival in HCC patients finding
  • GSTZ1 overexpression in HCC cells enriches metabolism/oxidative phosphorylation pathways and downregulates glycolysis and cancer pathways, suggesting a tumor-suppressive role mechanism
  • miR-539 is upregulated in HCC, negatively correlates with TAT/HPD/GSTZ1/FAH, and predicts worse survival, acting as an upstream regulator mechanism
  • Mutation and copy number alterations of tyrosine catabolic genes are infrequent and not strongly associated with their mRNA downregulation in HCC finding
  • Cross-platform integrated gene expression analysis (TCGA, GEO, GEPIA, Oncomine, KM plotter) provides a resource framework for identifying HCC biomarkers resource
Experimental setups
Assay System Perturbation Readout Platform
Pan-cancer transcriptome mRNA differential analysis (Oncomine) Multiple cancer types including HCC (LIHC), cancer vs normal tissue none Number of datasets with significant gene up/downregulation; gene rank percentile Oncomine database
RNA-seq transcriptome expression analysis (GEPIA) TCGA-LIHC plus GTEx; 369 HCC tissues vs 160 normal liver none log2(TPM+1) gene expression, one-way ANOVA GEPIA / TCGA / GTEx
Stage-wise gene expression (microarray/RNA-seq) GSE89377: normal (n=13), early HCC (n=5), stage 1 (n=9), stage 2 (n=12), stage 3 (n=14) none log2(TPM+1), Student's t-test GSE89377 (GEO)
Kaplan-Meier survival analysis TCGA-LIHC cohort of 364 liver cancer patients none Overall and disease-free survival by high vs low gene expression; HR, log-rank p GEPIA / KM plotter
Immunohistochemistry quantification HCC tumor vs normal liver tissue none Positive staining intensity fold-change, Student's t-test Human Protein Atlas (HPA)
Bulk RNA-seq differential expression and GSEA Huh7 HCC cell line GSTZ1 overexpression by adenoviral transfection vs control vector DEGs (DESeq2), pathway enrichment GSE117822; R/DESeq2; GSEA; Cytoscape Enrichment Map
Mutation and copy number analysis 353 HCC patients (TCGA) none Mutation type/SIFT impact, CNA frequency, GISTIC Q-value cBioPortal / GISTIC
miRNA target prediction and co-expression/survival analysis TCGA-LIHC (370 HCC, 50 normal); CapitalBio miRNA Array liver dataset none Predicted miRNA targets, miRNA-mRNA correlation (r), miRNA expression fold-change, survival HR TargetScan / starBase / KM plotter
Key results
  • TAT, HPD and GSTZ1 mRNA significantly decreased in HCC vs normal liver (TCGA/GEPIA); HGD and FAH virtually unchanged |Log2FC|=1, p<0.01
  • TAT, HPD, HGD, GSTZ1 and FAH transcripts significantly reduced in HCC stage 2 and 3 vs normal liver, but unchanged in early HCC p<0.05
  • Lower TAT, HGD and GSTZ1 expression associated with worse overall survival in HCC p=0.0067, p=0.0039, p=0.036
  • Lower TAT, HGD and GSTZ1 expression associated with worse disease-free survival p=0.011, p=0.0038, p=0.036
  • HCC tumor IHC staining reduced for HPD, HGD and GSTZ1 vs normal liver HPD 2.26-fold, HGD 1.67-fold, GSTZ1 2.27-fold
  • GSTZ1 overexpression in Huh7 yielded 3163 DEGs and downregulated glycolytic genes HK2 and PDK2 HK2 1.88-fold, PDK2 2.05-fold downregulated
  • miR-539 increased in HCC and negatively correlated with tyrosine catabolic gene expression 2.84-fold (p=0.05); r=-0.221, -0.193, -0.123, -0.166
  • High miR-539 expression led to worse overall survival in HCC patients
Key statistics
  • count 390, 428, 431, 457, 448 Oncomine datasets for TAT, HPD, HGD, GSTZ1, FAH (Pan-cancer Oncomine datasets meeting threshold (p≤1e-04, FC=2))
  • fold_change 2.26-fold (HPD), 1.67-fold (HGD), 2.27-fold (GSTZ1) decrease (IHC staining HCC vs normal; p=0.0388, 0.0423, 0.0007)
  • pvalue p=0.0067 (TAT), 0.0039 (HGD), 0.036 (GSTZ1) (Overall survival log-rank, TCGA-LIHC)
  • count 3163 DEGs (1742 up, 1421 down) (GSTZ1 overexpression vs control in Huh7, p<0.01)
  • fold_change miR-539 increased 2.84-fold (370 HCC vs 50 normal TCGA-LIHC, p=0.05)
  • correlation r=-0.221, -0.193, -0.123, -0.166 (miR-539 vs TAT, HPD, GSTZ1, FAH expression (starBase))
  • fold_change HK2 1.88-fold, PDK2 2.05-fold downregulated (Glycolytic genes in GSTZ1-overexpressing Huh7)
  • other each gene mutated in <1.1% of patients; HPD CNA Q-value=0.019 (Mutation/CNA profiles in TCGA HCC patients)

Statistical methods review

Model: opus

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 is a cross-platform, in silico bioinformatics study that re-analyzes publicly available transcriptomic, proteomic, mutation, and survival datasets (TCGA-LIHC, GEO, GEPIA, Oncomine, KM plotter, Human Protein Atlas, cBioPortal, starBase) to characterize five tyrosine catabolic genes in hepatocellular carcinoma. Differential expression between tumor and normal tissue was assessed with one-way ANOVA (GEPIA) and Student's t-tests (GEO/IHC quantification), prognosis with Kaplan-Meier curves and log-rank tests plus Cox regression, and pathway changes via DESeq2-derived DEGs followed by GSEA. Group differences are reported with significance thresholds and asterisk tiers, hazard ratios with 95% CIs for survival, and bar/box/violin summaries shown as mean ± SEM.

Replicationunclear Sample sizeSample counts reported per dataset/group (e.g., TCGA tumor n=369/normal n=160; survival n=364; GSE89377 per-stage n=5–14); no formal power/sample-size calculation described Groupstumor vs normal liver, across HCC stages, and high vs low expression subgroups Pairingunpaired Randomization/blindingna DispersionSEM Exact p-valuesyes Effect sizesyes Confidence intervalsyes Multiplicity correctionnone stated for the t-test/ANOVA comparisons; DESeq2 internally provides Benjamini-Hochberg FDR and GSEA provides FDR q-values, though the reported DEG and ANOVA thresholds are described as raw p-values
Statistical tests used
Test Applied to n Assumptions
one-way ANOVA (GEPIA differential analysis) tumor vs normal expression of tyrosine catabolic genes, Fig 2A (TCGA + GTEx) tumor n = 369, normal n = 160 not stated
Student's t-test (two-group, asterisk-tiered p<0.05/0.01/0.001) expression difference between tumor/stage and normal in GSE89377, Fig 2B normal n=13, early HCC n=5, stage1 n=9, stage2 n=12, stage3 n=14 not stated
Student's t-test IHC staining quantification tumor vs adjacent normal (HPD, HGD, GSTZ1, FAH), Fig 4 not stated
Kaplan-Meier with log-rank test overall survival and disease-free survival by high/low gene expression, Fig 3 and S2 (TCGA-LIHC, n=364) 364 patients na
Cox regression (proportional hazards) relapse-free survival prediction for miR-539 and miR-661, Fig 6B (KM plotter) not stated
DESeq2 Wald test for differential expression DEGs between GSTZ1-overexpressing and control Huh7 cells, GSE117822 (3163 DEGs, p<0.01) na
correlation analysis (Pearson/Spearman r, starBase) miR-539 vs target gene expression in HCC, S6 Fig (r = -0.221, -0.193, -0.123, -0.166) 370 HCC, 50 normal not stated
Approaches that could also have been used
  • Tumor-vs-normal and stage-vs-normal expression differences were compared with Student's t-test (and one-way ANOVA in GEPIA), with significance shown as asterisk tiers.
    Could also: A non-parametric test such as Mann-Whitney U (two groups) or Kruskal-Wallis (multiple stages) could also be used, and exact p-values could be tabulated alongside the asterisks. — Rank-based tests make fewer distributional assumptions, which can be informative for small per-stage groups (n = 5–14), and exact p-values convey the precise strength of evidence.
  • Several genes and stages were each compared to the normal group using separate t-tests.
    Could also: A single ANOVA across all stages followed by a post-hoc multiple-comparison correction (e.g., Tukey HSD or Dunnett's vs. control) could also frame these comparisons within one model. — A unified model with post-hoc correction also controls the family-wise error rate across the related comparisons and reports them together.
  • Dispersion in bar/box/violin summaries was reported as mean ± SEM.
    Could also: Standard deviation or a 95% confidence interval could also be reported. — SD or a CI directly conveys the spread of the data (rather than precision of the mean) and is often preferred, especially for small sample sizes.
  • DEG and ANOVA/t-test comparisons were reported against raw p-value thresholds (e.g., p<0.01).
    Could also: An explicitly stated FDR-adjusted threshold (e.g., Benjamini-Hochberg q-value) could also be applied and reported for the high-dimensional screens. — Reporting adjusted q-values also makes the false-discovery control explicit across the many genes/pathways tested simultaneously.
  • Survival subgroups were defined by dichotomizing expression at the median or quartile cut-points before log-rank testing.
    Could also: A Cox model treating expression as a continuous variable (optionally with covariate adjustment for stage/age) could also be fit. — Modeling expression continuously avoids the loss of information from dichotomization and can also adjust for potential confounders.
  • Correlations between miRNA and target gene expression were reported as r values.
    Could also: Reporting whether Pearson or Spearman was used, with 95% CIs on r, could also accompany the coefficients. — Specifying the method and adding CIs clarifies the linear vs. monotonic assumption and conveys the precision of the correlation estimate.
Software: R · DESeq2 · GEPIA (web) · Oncomine (web) · KM plotter (web) · cBioPortal / GISTIC · GSEA · Cytoscape / Enrichment Map · TargetScan / starBase (web)

Result convergence & founder nodes

Findings this paper shares with others that ran a comparable experiment. A node’s strength is how many independent papers report it (replication breadth) — not how often it is cited, so a heavily-replicated but under-cited founder still stands out.

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
54
Impact: high
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.

GSE117822 GEO in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
GSE89377 GEO in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
HPA004701 HPA in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
HPA038321 HPA in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
HPA041370 HPA in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
HPA047374 HPA in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet

Downstream reach in the literature

70 downstream papers · 2 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.

This paper is currently under reproducibility review (see the verdict above). The map below shows where the data in question has propagated — so reuse can be traced, not so the downstream work is presumed affected.

What was reproduced

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

Scope — pmid-32542016

Paper: Nguyen TN, Nguyen HQ, Le DH. Unveiling prognostics biomarkers of tyrosine metabolism reprogramming in liver cancer by cross-platform gene expression analyses. PLoS One 2020. PMID 32542016 / PMC7295234 / doi:10.1371/journal.pone.0229276.

Code: https://github.com/nguyenquyha/IHC-method (commit 41db5bbfe574d789a86181704c0adad5000055f7, branch master, pushed 2019-04-17). Data: GSE89377 (microarray, for the cross-platform DEG part) and the IHC images shipped inside the repo (FAH/GSTZ1/HGD/HPD folders, 112 PNGs + 112 masks).

What the repo actually is (P16: own code, ships data + expected output)

The repo is the authors' own IHC quantification method: a MATLAB script brown_calc.m that, for each immunohistochemistry image, builds a tissue mask (grayscale threshold → morphological open/close → largest connected component → convex hull) and counts the fraction of "brownish" pixels (DAB chromogen) inside that mask via an HSV color threshold. The repo ships:

  • input data: 112 IHC images (FAH 31, GSTZ1 15, HGD 32, HPD 34), each with its computed *_mask.png;
  • code: brown_calc.m (the full pipeline, self-contained);
  • expected output: brown_ratio_table.csv — one brown-ratio per image (112 rows).

This is the ideal reproduction unit: data + code + expected numeric output all shipped together, deterministic, no randomness.

In scope (pipeline-derived, attempted)

  1. R1 — brown_ratio_table.csv (primary, direct pipeline output). Re-run the brown-pixel quantification on the shipped images and compare all 112 per-image brown ratios to the shipped table. Two variants for auditability:
    • R1a (shipped-mask): compute the ratio using the shipped *_mask.png, isolating the deterministic color step (no morphology uncertainty).
    • R1b (full-pipeline): regenerate the mask from scratch (port of the MATLAB morphology) and recompute. Also compare regenerated mask vs shipped mask (Jaccard) to validate the morphology port.
  2. R2 — Figure 4 IHC fold-changes & t-test p-values (derived from R1). The paper reports decreased staining in tumor: GSTZ1 2.27-fold (p=0.0007), HGD 1.67-fold (p=0.0423), HPD 2.26-fold (p=0.0388); FAH not quantified in IHC. Fold = mean(N)/mean(T); unpaired Student t-test. Recompute from the reproduced ratios. (Aggregation of the "± x.xx" term is under-specified in the paper; noted.)

Out of scope (not attempted, and why)

  • Cross-platform microarray DEG / tyrosine-metabolism analysis on GSE89377 and the prognostic-biomarker / survival modeling (the paper's main bioinformatic thread): no analysis code is shipped for this part — the repo contains only the IHC image method. Reproducing it would require re-deriving an unspecified pipeline; per the 80/20 rule this is the hard, under-specified ~20% and is skipped (no_code for that sub-result, not the paper as a whole).
  • Wet-lab IHC staining itself (out of scope by definition — manual/experimental).

Compute

MATLAB is proprietary; the port is run in Python (numpy/scipy/scikit-image) on «our HPC» SLURM, matching MATLAB's rgb2gray coefficients, mat2gray global rescale, rgb2hsv, and disk-morphology as closely as feasible. R1a (shipped masks) is mathematically independent of the morphology port and is the high-confidence anchor.

Figures / tables: Fig 4Fig 4A
C1-brown_ratio_table
Reported
brown_ratio_table.csv: 112 per-image IHC brown-pixel ratios (4 enzymes), the direct deterministic pipeline output
Reproduced
R1a shipped-mask: max|d|=0.00182, mean|d|=6.3e-5, 110/112 within 1e-3, 112/112 within 1e-2, Pearson r=0.99999971; R1b full-pipeline Pearson 0.9999966; mask Jaccard mean 0.9978
within tolerance
C3-GSTZ1-fold
Reported
GSTZ1 IHC decreased 2.27-fold (p=0.0007), Fig 4
Reproduced
fold=2.269 (n_N=3,n_T=12), unpaired-t p=0.00068
exact
C4-HGD-fold
Reported
HGD IHC decreased 1.67-fold (p=0.0423), Fig 4
Reproduced
fold=2.051 (n_N=6,n_T=26), unpaired-t p=0.00026 (no aggregation variant recovers 1.67)
did not match
C5-HPD-fold
Reported
HPD IHC decreased 2.26-fold (p=0.0388), Fig 4
Reproduced
fold=1.983 (n_N=6,n_T=28), unpaired-t p=0.0065 (median 1.83; drop-near-zero-T 1.35; none reach 2.26)
did not match

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 51/100

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: Q6 · Severity of the deviation 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +5

The deterministic pipeline output (112-image brown_ratio_table.csv) reproduces essentially exactly (Pearson 0.99999971) and GSTZ1's Fig-4 fold and p reproduce to three decimals (2.269, p=0.00068), confirming both the data and the recipe are correct. Yet applying that identical recipe to HGD (1.67) and HPD (2.26) yields 2.05 and 1.98 — not recoverable from the shipped table by any tested aggregation, with reported borderline p~0.04 vs computed p<0.01. This sits on the authors' side (a likely different image subset/version or transcription slip behind those panels; the repo also ships FAH data the paper says it didn't quantify), not on our method, and is moderate (direction and significance hold). Overall a solid, well-anchored reproduction with two clearly flagged, non-derivable figure values; the paper's main cross-platform analysis was out of scope (no code).

🤝
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 support@doesitreproduce.com.

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.

138.3 k
tokens (I/O) · 7.8 M incl. cache
13 min
runtime · 0.04 CPU-h
3.3 GB
peak RAM
1
HPC jobs
hummel
machine