Model

scGPT

scGPT learns representations of single-cell molecular measurements and supports task-specific adaptation.

Sources (3)

bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row

3 evaluations · 3 metric rows · 3 evaluated configurations using this model

How it worksscGPT workflow

Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation.

Sources (3)

Overview

Model type

Generative single-cell transformer

Sources (3)

Inputs

Gene-expression measurements with the checkpoint-matched gene vocabulary.

Sources (3)

Outputs

Cell/gene representations and task-specific predictions after the relevant workflow.

Sources (3)

Access

Official project documentation and implementation: https://github.com/bowang-lab/scGPT

Sources (3)

limited source coverage · Automated source review, 2026-09-16. All specifications and missing details

Evaluations and results

3 evaluations · 3 metric rows. Different protocols are not a single leaderboard.

Filter evaluations

Tested model or methodProtocolDatasetEvaluation setupMetricEvidence origin

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configuration	Protocol and dataset	Finding	Evidence and details
Configuration: scGPT	Task: Cell-type identification Dataset: M.S. single-cell dataset	0.734 F1-Score unitless · unknown Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible	Independent external evaluation · source checked Methods, coverage and source scGPT: Cell-type identification Native scLLM cell-type identification as reported in Table 2. Aggregation: Not reported Parameter-Efficient Fine-Tuning Enhances Adaptation of Single Cell Large Language Model for Cell Type Identification · Table 2, M.S. / scGPT row, F1-Score column
Configuration: scGPT	Task: Cell-type annotation Dataset: hPancreas	0.550 F1 unitless · unknown Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible	Result quoted from another source · source checked Methods, coverage and source scGPT: Cell-type annotation Zero-shot setting; source caption says some comparator rows come from GenePT. Aggregation: Not reported scELMo: Embeddings from Language Models are Good Learners for Single-cell Data Analysis · Table 1, hPancreas zero-shot / scGPT (z) row, F1 column
Configuration: scGPT	Task: Cell-type structure in frozen embeddings Dataset: Aorta single-cell dataset	0.47 Adjusted Rand Index unitless · unknown Uncertainty: Not reported Coverage: Not reported scored / Not reported eligible	Independent external evaluation · source checked Methods, coverage and source scGPT: Cell-type structure in frozen embeddings k-means on pretrained cell embeddings; agreement with original cell-type labels. Aggregation: Not reported GenePT: A Simple But Effective Foundation Model for Genes and Cells Built From ChatGPT · Table 2, Aorta / Cell type row, scGPT ARI column

Source checking is not independent reproduction. Release 2026-09-23-2b89723c6dd9.

Related configurations, pipelines and services

These configurations, services and pipelines use this model within their own configurations. Their results, where available, are not assigned to the underlying model.

CPA* (scGPT) Configuration · 8 metric rows
LA (scGPT) Configuration · 8 metric rows
scGPT + residual geometry Pipeline · 1 metric rows

Use this model

How it works, versions and access

Related profile: scGPT. This page retains the exact record and its evaluation context.

Versions and evaluated configurations

scGPT · alias of
scGPT · family
scGPT · family
scGPT · family

How it works

scGPT combines each gene identity with its expression-value encoding before transformer attention. The implementation supports several expression encoders and cell-pooling choices. Task heads then predict expression or cell labels; optional masking and batch objectives depend on the training configuration.

Sources (3)

Versions and reproducibility

The May 2023 preprint reports an early 10M-cell model. The current whole-human checkpoint table reports 33M normal human cells; these sources describe different releases. The applicable input limits require configuration-specific checking.

Sources (3)

Strengths, limitations and unresolved questions

Strengths and limitations

Strengths and considerations

Releases include whole-human and organ-specific checkpoints, with tutorials for reference mapping and other downstream tasks.
Sources (3)
bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row

Limitations and conditions

Checkpoint choice and vocabulary must match the biological context. A whole-human pretrained encoder and a fine-tuned annotation or perturbation model are distinct evaluated configurations.
Sources (3)
bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row

What remains unknown

Parameters: The May 2023 model has 12 transformer blocks, width 512 and eight heads. The inspected current model-zoo table does not state the exact parameter total of its separate 33M-cell checkpoint.
Training cutoff: The reviewed early manuscript and current whole-human model-zoo entry describe different corpora; neither supplies a shared latest-study date for the current checkpoint.
Context limits: The implementation accepts variable gene sets matched to its vocabulary. The reviewed model-zoo entry does not specify one validated maximum gene sequence for the current whole-human checkpoint.
Weights licence: Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant.

Profile review details

Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied.

Stable record: catalog-model-scgpt

Specifications

Inputs, training, access and other details

Explanatory profile: limited source coverage · Automated source review, 2026-09-16. Review applies to the cited claims; unresolved fields are listed below. Numerical results retain their own review status.

Inputs, outputs and configuration
Property	Description and evidence
Model type	Generative single-cell transformer Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Architecture	Transformer backbone combining learned gene-token embeddings with expression-value encodings and optional batch encodings. Separate expression, cell-classification and optional masked-value or batch-discriminator heads support configured tasks. Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Inputs	Gene-expression measurements with the checkpoint-matched gene vocabulary. Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Outputs	Cell/gene representations and task-specific predictions after the relevant workflow. Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Parameters	The May 2023 model has 12 transformer blocks, width 512 and eight heads. The inspected current model-zoo table does not state the exact parameter total of its separate 33M-cell checkpoint. · Not reported in inspected sources Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Known versions	The May 2023 preprint reports an early 10M-cell model. The current whole-human checkpoint table reports 33M normal human cells; these sources describe different releases. Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Training data	The current whole-human model-zoo checkpoint uses 33M normal human cells, alongside separately released organ-specific and pan-cancer models. The earlier May 2023 preprint describes 10M training cells; its corpus is not the current checkpoint corpus. Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Training cutoff	The reviewed early manuscript and current whole-human model-zoo entry describe different corpora; neither supplies a shared latest-study date for the current checkpoint. · Not reported in inspected sources Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Context limits	The implementation accepts variable gene sets matched to its vocabulary. The reviewed model-zoo entry does not specify one validated maximum gene sequence for the current whole-human checkpoint. · Not reported in inspected sources Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Weights licence	Separate checkpoint-distribution terms are not stated in the inspected release documentation and licence material. The source-code licence alone is not recorded as an explicit weight grant. · Not reported in inspected sources Sources (4) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF; bowang-lab/scGPT: LICENSE · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row; LICENSE: licence text
Access	Official project documentation and implementation: https://github.com/bowang-lab/scGPT Sources (3) bowang-lab/scGPT: README.md; bowang-lab/scGPT: scgpt/model/model.py; scgpt-may2023: Primary paper PDF · May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row
Code licence	MIT Sources bowang-lab/scGPT: LICENSE · LICENSE: licence text

Evidence

Source checking verifies the cited claim or transcription. It does not establish 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.

60 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-23-2b89723c6dd9
Property and statement	Original source and location	Review and provenance
Diagram caption Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation. Individual claims	bowang-lab/scGPT: README.md Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram caption Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation. Individual claims	bowang-lab/scGPT: scgpt/model/model.py Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `4ae77618cc6f12a7d1f6d946fee3f7bd5c56f3e9e103073394a590ceba5a35ed` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram caption Conceptual summary of the documented data flow; optional inputs and configured downstream stages must be reported for a reproducible evaluation. Individual claims	scgpt-may2023: Primary paper PDF Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: bioRxiv version posted 1 May 2023; re-inspected existing local research artifact Retrieved: 2026-09-16T20:21:23.046376+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `4ef64d2b3431812df07c4b3e428950644ff264ec5728267c2e4692588f557419` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram steps Gene expression and vocabulary scGPT encoder Cell and gene representations Task-specific adaptation Individual claims	bowang-lab/scGPT: README.md Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram steps Gene expression and vocabulary scGPT encoder Cell and gene representations Task-specific adaptation Individual claims	bowang-lab/scGPT: scgpt/model/model.py Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `4ae77618cc6f12a7d1f6d946fee3f7bd5c56f3e9e103073394a590ceba5a35ed` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram steps Gene expression and vocabulary scGPT encoder Cell and gene representations Task-specific adaptation Individual claims	scgpt-may2023: Primary paper PDF Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: bioRxiv version posted 1 May 2023; re-inspected existing local research artifact Retrieved: 2026-09-16T20:21:23.046376+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `4ef64d2b3431812df07c4b3e428950644ff264ec5728267c2e4692588f557419` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram title scGPT workflow Individual claims	bowang-lab/scGPT: README.md Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram title scGPT workflow Individual claims	bowang-lab/scGPT: scgpt/model/model.py Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `4ae77618cc6f12a7d1f6d946fee3f7bd5c56f3e9e103073394a590ceba5a35ed` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Diagram title scGPT workflow Individual claims	scgpt-may2023: Primary paper PDF Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: bioRxiv version posted 1 May 2023; re-inspected existing local research artifact Retrieved: 2026-09-16T20:21:23.046376+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `4ef64d2b3431812df07c4b3e428950644ff264ec5728267c2e4692588f557419` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact
Model type Generative single-cell transformer Individual claims	bowang-lab/scGPT: README.md Original source ↗ May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: cebd6fae655b9c585a4807daa3ac31bb764f06b4 Retrieved: 2026-09-16T19:46:18.594429+00:00	source checked automated source review · 2026-09-16 Audit details Inspected pinned official documentation, relevant implementation files and named primary-paper sections. Claims are limited to those artifacts. Remaining field extraction and identity conflicts are explicit; no new performance claims, model runs or human review are implied. Field: `attributes.profile.facts.0.value` Source artifact SHA-256: `b0503e8ca789f19f1fc2350c5aaf57b1b323bbae43b354655231b5f4a1586c83` Hash scope: SHA-256 of retrieved original artifact bytes Format: original_artifact Inspected artifact

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

Sources and history

View linked audit checks and correction history

Release 2026-09-23-2b89723c6dd9 · Record review: discovered

5 source records and release history

scGPT official resource · Original source · Not reported
bowang-lab/scGPT: README.md · Original source · cebd6fae655b9c585a4807daa3ac31bb764f06b4
bowang-lab/scGPT: scgpt/model/model.py · Original source · cebd6fae655b9c585a4807daa3ac31bb764f06b4
scgpt-may2023: Primary paper PDF · Original source · bioRxiv version posted 1 May 2023; re-inspected existing local research artifact
bowang-lab/scGPT: LICENSE · Original source · cebd6fae655b9c585a4807daa3ac31bb764f06b4

Download this release

Technical metadata and extraction receipts

Stable ID: catalog-model-scgpt

areas: cells-tissues
method types: foundation model
entity level: family
version: whole-human
reported name: scGPT
access: Public code and downloadable checkpoints; use the unfine-tuned whole-human model for a new task.
method type: foundation model
historical missing metadata: checkpoint revision: not_yet_extracted; training data: not_yet_extracted; licence: not_yet_extracted
metadata review scope: historical_missing_metadata preserves the original discovery state. Current descriptive evidence and missingness are recorded in profile.facts; numerical-result review is separate.
entity classification: review date: 2026-09-17; rationale: The cited profile describes a named learned biological predictor or representation model/family. Preserve this identity separately from task-specific fitting, individual checkpoints, pipelines and hosted access.; source ids: evidence-official-0e4dac85aa0e0ff1d4b2; evidence-official-d8e2d4f2e752370b6565; evidence-official-efa6ec9d886a21534a20; source locator: May 2023 scGPT preprint Sections 4.1–4.3 and 4.8; current scgpt/model/model.py and README pretrained model table; README.md: Pretrained scGPT Model Zoo, whole-human row; ambiguities: None recorded

Related records

variant of: scGPT
alias of: scGPT
uses model: CPA* (scGPT)
uses model: LA (scGPT)
subject: scGPT: variant of scGPT
family: scGPT
family: scGPT
family: scGPT
uses model: scGPT + residual geometry

Suggest a correction