Norman19 (PerturBench split)

Subset and evaluation context

This record describes a particular subset or cohort used in an evaluation. Its results do not describe the full dataset.

Evaluation results

Release 2026-09-17-134cd1815de8 · 36 evaluations · 36 metric rows. Different protocols are not a single leaderboard. Where several source tables report the same metric, the published comparisons above offer a pooled view that names what it does not hold constant.

Evaluation setupMetricEvidence origin

Results grouped by the exact reported evaluation
Metric and finding	Coverage and uncertainty	Evidence
Biolord* on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: Biolord*Task: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.41 ± 2 × 10 − 2 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.02 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Biolord ∗), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.
Biolord* on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank Configuration: Biolord*Task: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.027 ± 1 × 10 − 3 cosine_logfc_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.001 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Biolord ∗), column(Cosine, LogFC rank) Source checking is not independent reproduction.
Biolord* on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank Configuration: Biolord*Task: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.028 ± 1 × 10 − 3 rmse_mean_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.001 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Biolord ∗), column(RMSE, mean rank) Source checking is not independent reproduction.
Biolord* on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean Configuration: Biolord*Task: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the meanDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.086 ± 6 × 10 − 4 rmse_mean Unit: error · Direction: lower	Uncertainty: type: standard deviation; value: 0.0006 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Biolord ∗), column(RMSE, mean) Source checking is not independent reproduction.
CPA* on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: CPA*Task: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.52 ± 6 × 10 − 2 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.06 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.
CPA* on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank Configuration: CPA*Task: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.12 ± 2 × 10 − 2 cosine_logfc_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.02 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗), column(Cosine, LogFC rank) Source checking is not independent reproduction.
CPA* on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank Configuration: CPA*Task: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.17 ± 3 × 10 − 2 rmse_mean_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.03 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗), column(RMSE, mean rank) Source checking is not independent reproduction.
CPA* on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean Configuration: CPA*Task: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the meanDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.079 ± 5 × 10 − 3 rmse_mean Unit: error · Direction: lower	Uncertainty: type: standard deviation; value: 0.005 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗), column(RMSE, mean) Source checking is not independent reproduction.
CPA* (noAdv) on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: CPA* (noAdv)Task: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.55 ± 9 × 10 − 2 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.09 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (noAdv)), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.
CPA* (noAdv) on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank Configuration: CPA* (noAdv)Task: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.16 ± 4 × 10 − 2 cosine_logfc_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.04 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (noAdv)), column(Cosine, LogFC rank) Source checking is not independent reproduction.
CPA* (noAdv) on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank Configuration: CPA* (noAdv)Task: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.18 ± 2 × 10 − 2 rmse_mean_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.02 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (noAdv)), column(RMSE, mean rank) Source checking is not independent reproduction.
CPA* (noAdv) on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean Configuration: CPA* (noAdv)Task: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the meanDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.073 ± 6 × 10 − 3 rmse_mean Unit: error · Direction: lower	Uncertainty: type: standard deviation; value: 0.006 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (noAdv)), column(RMSE, mean) Source checking is not independent reproduction.
CPA* (scGPT) on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: CPA* (scGPT)Task: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.70 ± 1 × 10 − 2 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.01 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.
CPA* (scGPT) on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank Configuration: CPA* (scGPT)Task: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.064 ± 1 × 10 − 2 cosine_logfc_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.01 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(Cosine, LogFC rank) Source checking is not independent reproduction.
CPA* (scGPT) on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank Configuration: CPA* (scGPT)Task: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.13 ± 2 × 10 − 2 rmse_mean_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.02 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(RMSE, mean rank) Source checking is not independent reproduction.
CPA* (scGPT) on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean Configuration: CPA* (scGPT)Task: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the meanDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.061 ± 2 × 10 − 3 rmse_mean Unit: error · Direction: lower	Uncertainty: type: standard deviation; value: 0.002 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(CPA ∗ (scGPT)), column(RMSE, mean) Source checking is not independent reproduction.
Decoder on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: DecoderTask: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.73 ± 2 × 10 − 2 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.02 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Decoder), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.
Decoder on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank Configuration: DecoderTask: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.017 ± 6 × 10 − 3 cosine_logfc_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.006 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Decoder), column(Cosine, LogFC rank) Source checking is not independent reproduction.
Decoder on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank Configuration: DecoderTask: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.014 ± 4 × 10 − 4 rmse_mean_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.0004 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Decoder), column(RMSE, mean rank) Source checking is not independent reproduction.
Decoder on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean Configuration: DecoderTask: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the meanDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.043 ± 3 × 10 − 4 rmse_mean Unit: error · Direction: lower	Uncertainty: type: standard deviation; value: 0.0003 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(Decoder), column(RMSE, mean) Source checking is not independent reproduction.
LA on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: LATask: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.79 ± 1 × 10 − 2 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.01 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(LA), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.
LA on PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rank Configuration: LATask: PerturBench CB-COSINE-RANK: combination prediction on Norman19, Cosine LogFC rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.005 ± 2 × 10 − 3 cosine_logfc_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.002 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(LA), column(Cosine, LogFC rank) Source checking is not independent reproduction.
LA on PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rank Configuration: LATask: PerturBench CB-RMSE-RANK: combination prediction on Norman19, RMSE mean rankDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.014 ± 1 × 10 − 3 rmse_mean_rank Unit: fraction · Direction: lower	Uncertainty: type: standard deviation; value: 0.001 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(LA), column(RMSE, mean rank) Source checking is not independent reproduction.
LA on PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the mean Configuration: LATask: PerturBench CB-RMSE: combination prediction on Norman19, RMSE of the meanDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.043 ± 4 × 10 − 4 rmse_mean Unit: error · Direction: lower	Uncertainty: type: standard deviation; value: 0.0004 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(LA), column(RMSE, mean) Source checking is not independent reproduction.
LA (scGPT) on PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold change Configuration: LA (scGPT)Task: PerturBench CB-COSINE: combination prediction on Norman19, Cosine similarity of log fold changeDataset subset: Norman19 (PerturBench split) Predict the effect of a pair of gene overexpressions from single perturbations, reported as the mean and one standard deviation over seeds. Author-reported evaluation · Evaluation metadata: source checked
0.77 ± 4 × 10 − 3 cosine_logfc Unit: fraction · Direction: higher	Uncertainty: type: standard deviation; value: 0.004 Scored: Not reported · Eligible: Not reported	source checkedPerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Table 3, row(LA (scGPT)), column(Cosine, log fold change (LogFC)) Source checking is not independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

2 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-17-134cd1815de8
Property and statement	Original source and location	Review and provenance
description The split of Norman19 that PerturBench evaluated on. The upstream dataset release is not catalogued here, so no claim is made that this matches its original splits. Context-only references	PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis Original source ↗ No field-specific location recorded Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256 Retrieved: 2026-09-17T08:06:28.400469+00:00	not individually reviewed No individual claim review recorded Audit details Field: `description` Source artifact SHA-256: `5c4804565dd9faa17a11853a79e9847dcb6da73715b4c91f62e5b274cc79f186` Hash scope: Exact retrieved primary paper artifact bytes. Inspected artifact
name Norman19 (PerturBench split) Context-only references	PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis Original source ↗ No field-specific location recorded Version: Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256 Retrieved: 2026-09-17T08:06:28.400469+00:00	not individually reviewed No individual claim review recorded Audit details Field: `name` Source artifact SHA-256: `5c4804565dd9faa17a11853a79e9847dcb6da73715b4c91f62e5b274cc79f186` Hash scope: Exact retrieved primary paper artifact bytes. Inspected artifact

All evidence (CSV)All evidence (JSONL)How to read the evidence

Sources and history

Release 2026-09-17-134cd1815de8 · Record review: source checked

1 source records and release history

PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis · Original source · Primary full-text snapshot retrieved 2026-09-17; exact bytes pinned by SHA-256

Download this release