Configuration

GREmLN

GREmLN learns single-cell representations using a gene-interaction graph inside its attention mechanism.

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Abstract (paragraph 1); Validation Experiments/Graph Structure Understanding (paragraph 3)

1 evaluation · 1 metric row

How it worksEvaluated procedure (conceptual)

Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Model Architecture (paragraph 1); Abstract (paragraph 1)

At a glance

Model type

Graph-based predictive method; this record is the paper-specific evaluated configuration.

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Model Architecture (paragraph 1); Abstract (paragraph 1)

Biological inputs

Single-cell expression profiles and a molecular-interaction graph

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Validation Experiments/Transcriptomic Landscape Learning & Cell Type Annotation/Bayesian Graph Integration (paragraph 3); Model Architecture (paragraph 1)

Outputs

Gene/cell embeddings and adapted cell-annotation or reverse-perturbation predictions

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Abstract (paragraph 1); Validation Experiments/Fine-tuning for Reverse Perturbation Prediction (paragraph 1)

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

Evaluations and results

Release 2026-09-17-d277315f7d76 · 1 evaluation · 1 metric row. Different protocols are not a single leaderboard.

Evaluation setupMetricEvidence origin

Results grouped by the exact reported evaluation
Metric and finding	Coverage and uncertainty	Evidence
GREmLN: cell-type annotation Configuration: GREmLNTask: cell-type annotationDataset: non-immune cells Zero-shot cell-type annotation using pre-trained cellular graph foundation model Author-reported evaluation · Evaluation metadata: needs review
0.937 F1 Unit: fraction · Direction: unknown	Uncertainty: not reported in legacy extract Scored: Not reported · Eligible: Not reported	source checkedGREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Table 2, Cell type annotation(zero-shot), Non-immune cells, F1 row, GREmLN column Source checking is not independent reproduction.

How it works

How the evaluated method works

Graph signal processing supplies gene-token structure to attention, replacing an arbitrary sequence ordering with regulatory or interaction relationships.

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Model Architecture (paragraph 1); Abstract (paragraph 1)

What was evaluated

The linked evaluation record identifies GREmLN: cell-type annotation. Its dataset, split, adaptation and evidence origin remain attached to the reported results.

Sources

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b4-014

Strengths and limitations

Strengths and considerations

Makes graph relationships explicit in the representation of unordered gene-expression features.
Sources
GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Validation Experiments/Graph Ablation (paragraph 1); Conclusion & Discussion (paragraph 1)

Limitations and conditions

The graph is an additional information source; comparisons with expression-only models must account for its provenance and coverage.
Sources
GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Validation Experiments/Graph Ablation (paragraph 1); Appendix/A Appendix/A.1 Gene Regulatory Network Construction (paragraph 1)

Profile review details

Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction.

Stable record: reported-model-53d6515bcc1f39

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	Graph-based predictive method; this record is the paper-specific evaluated configuration. Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Model Architecture (paragraph 1); Abstract (paragraph 1)
Architecture / procedure	Graph signal processing supplies gene-token structure to attention, replacing an arbitrary sequence ordering with regulatory or interaction relationships. Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Model Architecture (paragraph 1); Abstract (paragraph 1)
Biological inputs	Single-cell expression profiles and a molecular-interaction graph Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Validation Experiments/Transcriptomic Landscape Learning & Cell Type Annotation/Bayesian Graph Integration (paragraph 3); Model Architecture (paragraph 1)
Outputs	Gene/cell embeddings and adapted cell-annotation or reverse-perturbation predictions Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Abstract (paragraph 1); Validation Experiments/Fine-tuning for Reverse Perturbation Prediction (paragraph 1)
Parameters	The principal model is reported as 10.3M learnable parameters. A separate scaling study uses one-, three- and six-layer models reported as 7.4M, 10M and 24.2M; these are distinct configurations. Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Validation Experiments/Scaling Behavior (paragraph 1); Appendix/A Appendix/A.6 Computational Costs/Pretraining (paragraph 1)
Known versions / configuration	GREmLN is the comparison-table label; that label does not specify an immutable weight revision. · Not reported in inspected sources Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label.
Training data / fitting	Masked expression-bin prediction on a CELLxGENE-derived corpus with cell-type-specific gene regulatory graphs. Appendix A.2 holds out 15% of cell types/graphs and leaves about 70% of cells for training; Appendix A.5 describes one pretraining epoch. Sources GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Appendix/A Appendix/A.9 Details of Validation Datasets/Human Immune Cells (paragraph 1); Model Architecture/Training Objectives (paragraph 1)
Context limits	The pinned implementation’s example tokenizer uses max_seq_length=4096. This is an explicit documented configuration, not evidence that every historical evaluation used the same limit. Sources czi-ai/GREmLN README.md · README.md; Custom Tokenization Example
Access	Official study implementation and usage documentation: https://github.com/czi-ai/GREmLN/blob/e1c5d8edbe2fe96568ff5451f15bc691722bb6a1/README.md. This pinned documentation revision is not automatically the evaluated weight revision. Sources czi-ai/GREmLN README.md · README.md; installation, model download and usage instructions
Code licence	MIT (study repository code at the cited revision; this does not establish every dependency or historical checkpoint licence). Sources czi-ai/GREmLN LICENSE.md · LICENSE.md; complete licence text
Weights licence	The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately. · Not reported in inspected sources Sources czi-ai/GREmLN README.md · README.md; checkpoint/access documentation and licence scope

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.

19 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-09-17-d277315f7d76
Property and statement	Original source and location	Review and provenance
Diagram caption Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings. Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Model Architecture (paragraph 1); Abstract (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.diagram.caption` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram steps ["Single-cell expression profiles and a molecular-interaction graph","GREmLN","Gene/cell embeddings and adapted cell-annotation or reverse-perturbation predictions"] Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Model Architecture (paragraph 1); Abstract (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.diagram.steps` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram title Evaluated procedure (conceptual) Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Model Architecture (paragraph 1); Abstract (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.diagram.title` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Model type Graph-based predictive method; this record is the paper-specific evaluated configuration. Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Model Architecture (paragraph 1); Abstract (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.0.value` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Architecture / procedure Graph signal processing supplies gene-token structure to attention, replacing an arbitrary sequence ordering with regulatory or interaction relationships. Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Model Architecture (paragraph 1); Abstract (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.1.value` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Weights licence The inspected model-access documentation does not explicitly identify terms for this exact evaluated checkpoint or fitted head; repository code terms are shown separately. Individual claims	czi-ai/GREmLN README.md Original source ↗ README.md; checkpoint/access documentation and licence scope Version: e1c5d8edbe2fe96568ff5451f15bc691722bb6a1 Retrieved: 2026-09-16T20:42:56.518821+00:00	unreported automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.10.value` Source artifact SHA-256: `e6365c6df93ad707c57e04d57b6e3aa4121a1d36672e9a4f75db4a8ac664a858` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Biological inputs Single-cell expression profiles and a molecular-interaction graph Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Validation Experiments/Transcriptomic Landscape Learning & Cell Type Annotation/Bayesian Graph Integration (paragraph 3); Model Architecture (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.2.value` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Outputs Gene/cell embeddings and adapted cell-annotation or reverse-perturbation predictions Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Abstract (paragraph 1); Validation Experiments/Fine-tuning for Reverse Perturbation Prediction (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.3.value` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Parameters The principal model is reported as 10.3M learnable parameters. A separate scaling study uses one-, three- and six-layer models reported as 7.4M, 10M and 24.2M; these are distinct configurations. Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Validation Experiments/Scaling Behavior (paragraph 1); Appendix/A Appendix/A.6 Computational Costs/Pretraining (paragraph 1) Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	source checked automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.4.value` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Known versions / configuration GREmLN is the comparison-table label; that label does not specify an immutable weight revision. Individual claims	GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model Original source ↗ Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label. Version: preprint version in PMC Retrieved: 2026-09-16T10:33:57.502Z	unreported automated source review · 2026-09-16 Audit details Primary full text and the available official implementation/model documentation were inspected. Explanatory claims are source-backed; unresolved exact-configuration metadata is labelled explicitly. This is automated review, not a human review or independent benchmark reproduction. Field: `attributes.profile.facts.5.value` Source artifact SHA-256: `3a20c4ededb749fc3f1120baf16dcfebe3fcb30418a91c445cfd91a7b5fdf553` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact

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

Sources and history

Release 2026-09-17-d277315f7d76 · Record review: needs review

3 source records and release history

GREmLN: A Cellular Graph Structure Aware Transcriptomics Foundation Model · Original source · preprint version in PMC
czi-ai/GREmLN README.md · Original source · e1c5d8edbe2fe96568ff5451f15bc691722bb6a1
czi-ai/GREmLN LICENSE.md · Original source · e1c5d8edbe2fe96568ff5451f15bc691722bb6a1

Download this release

Technical metadata and extraction receipts

Stable ID: reported-model-53d6515bcc1f39

areas: cells-tissues
entity level: method
version: Not reported
reported name: GREmLN
historical missing metadata: version: not_reported_in_legacy_extract; checkpoint revision: not_reported_in_legacy_extract; training data: not_reported_in_legacy_extract; licence: not_reported_in_legacy_extract
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.
legacy kinds: model
entity classification: review date: 2026-09-17; rationale: This source-scoped entry preserves the method/configuration actually named in an evaluation. It is neither a global family identity nor proof of an immutable checkpoint; the linked evaluation retains adaptation, fitting and scoring details.; source ids: gremln-2026; source locator: Model Architecture (paragraph 1); Abstract (paragraph 1) | Abstract (paragraph 1); Validation Experiments/Graph Structure Understanding (paragraph 3); ambiguities: Configuration means the source-labelled evaluated identity. It does not establish missing checkpoint hashes, default settings or equivalence to same-named records in other papers.

Related records

model: GREmLN: cell-type annotation

Suggest a correction