Configuration

DNABERT-2

DNABERT-2 is fine-tuned to distinguish natural from GenomeOcean-generated DNA sequences.

Sources

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/Model Safety (paragraph 1); Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4)

1 evaluation · 3 metric rows

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

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2)

At a glance

Model type

DNA sequence transformer; this record is the paper-specific evaluated configuration.

Sources

MAGICS-LAB/DNABERT_2 README.md · README.md model description

Biological inputs

2-kbp natural or generated DNA sequences

Sources

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4); Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1)

Outputs

Natural-versus-artificial sequence predictions

Sources

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1); Results/Model Safety (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 · 3 metric rows. Different protocols are not a single leaderboard.

Evaluation setupMetricEvidence origin

Results grouped by the exact reported evaluation
Metric and finding	Coverage and uncertainty	Evidence
DNABERT-2: Natural vs artificial microbial genome sequence Configuration: DNABERT-2Protocol: Natural versus GenomeOcean-generated DNA classification (Natural vs artificial microbial genome sequence)Dataset: GenomeOcean natural/artificial sequence test DNABERT 2 and NTv 2 standard fine-tuning; GenomeOcean LoRA. Negatives generated by GenomeOcean itself. CAMI 2 train 18,000/validation 2,000; GTDB test 20,000; balanced natural/artificial 2 kb sequences. Independent external evaluation · Evaluation metadata: needs review
85.12 F1 Unit: % · Direction: higher	Uncertainty: not reported in legacy extract Scored: Not reported · Eligible: Not reported	source checkedGenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies; GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Table 2, DNABERT-2 row, F1 column Source checking is not independent reproduction.
85.02% Recall Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedGenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Table 2:, row DNABERT-2, column Recall; XML row2 column3 Source checking is not independent reproduction.
85.23% Precision Unit: percent · Direction: higher	Uncertainty: unreported Scored: Not reported · Eligible: Not reported	source checkedGenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Table 2:, row DNABERT-2, column Precision; XML row2 column2 Source checking is not independent reproduction.

How it works

How the evaluated method works

The pretrained DNA transformer is adapted with a binary classifier using the study’s natural/artificial sequence labels.

Sources

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2)

Underlying method and version boundaries

DNABERT-2 replaces overlapping k-mer tokens with byte-pair encoding and uses ALiBi positional biases. The official 117M model produces 768-dimensional token representations; downstream classifiers and pooling choices are separate configuration details.

Sources

MAGICS-LAB/DNABERT_2 README.md · README.md; introduction, model description, pretrained-model and usage sections at pinned revision

What was evaluated

The linked evaluation record identifies DNABERT-2: Natural vs artificial microbial genome sequence. Its dataset, split, adaptation and evidence origin remain attached to the reported results.

Sources

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · The named evaluation’s methods and comparison table; exact preserved evaluation IDs: evaluation-lit-b3-028

Strengths and limitations

Strengths and considerations

Tests whether generated DNA can be distinguished by a separate pretrained genome model.
Sources
GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/Model Safety (paragraph 1); Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1)

Limitations and conditions

A detector for this generator and data distribution is not a universal provenance or sequence-safety detector.
Sources
GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/Model Safety (paragraph 2); Discussion (paragraph 2)

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-7f1165b35f10e2

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	DNA sequence transformer; this record is the paper-specific evaluated configuration. Sources MAGICS-LAB/DNABERT_2 README.md · README.md model description
Architecture / procedure	The pretrained DNA transformer is adapted with a binary classifier using the study’s natural/artificial sequence labels. Sources GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2)
Biological inputs	2-kbp natural or generated DNA sequences Sources GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4); Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1)
Outputs	Natural-versus-artificial sequence predictions Sources GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1); Results/Model Safety (paragraph 1)
Parameters	An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources. · Not reported in inspected sources Sources (2) GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies; MAGICS-LAB/DNABERT_2 README.md · Results/Pre-training; Methods/Model and Training/Tokenization; Methods/Model and Training/Architecture and Pre-training; Methods/Model and Training/Fine-tune GenomeOcean as Biosynthetic Gene Clusters Foundation Model (bgcFM); Methods/Training/Evaluation Datasets/Assembled Metagenome Datasets; Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity; Methods/Training/Evaluation Datasets/Biosynthetic Gene Cluster (BGC) Collection; Methods/Training/Evaluation Datasets/ZymoBiomics Microbial Community Standard Datasets; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision
Known versions / configuration	DNABERT-2 is the comparison-table label; that label does not specify an immutable weight revision. · Not reported in inspected sources Sources GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Model identification in the comparison table and corresponding Methods; immutable checkpoint revision is not supplied by the table label.
Training data / fitting	Balanced 18,000/2,000/20,000 train/validation/test sequences. Sources GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1); Data & Code/Metagenome Raw Reads Datasets Used in Assembly (paragraph 1)
Context limits	2,000-bp benchmark windows. Sources GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Abstract (paragraph 1); Table T4 (paragraph 1)
Access	Official upstream implementation and usage documentation: https://github.com/MAGICS-LAB/DNABERT_2/blob/f25bed9ee20db966dff39e5c1571249d04e36404/README.md. This pinned documentation revision is not automatically the evaluated weight revision. Sources MAGICS-LAB/DNABERT_2 README.md · README.md; installation, model download and usage instructions
Code licence	Apache 2.0 (upstream repository code at the cited revision; this does not establish every dependency or historical checkpoint licence). Sources MAGICS-LAB/DNABERT_2 LICENSE · LICENSE; 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 MAGICS-LAB/DNABERT_2 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.

21 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	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2) Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram steps ["2-kbp natural or generated DNA sequences","DNABERT-2","Natural-versus-artificial sequence predictions"] Individual claims	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2) Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Diagram title Evaluated procedure (conceptual) Individual claims	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2) Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Model type DNA sequence transformer; this record is the paper-specific evaluated configuration. Individual claims	MAGICS-LAB/DNABERT_2 README.md Original source ↗ README.md model description Version: f25bed9ee20db966dff39e5c1571249d04e36404 Retrieved: 2026-09-16T20:00:02.624261+00:00	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: `734a8cec5f667d74d421bf3b273ad7e256216109636da45aa7ceba21cd34de16` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Architecture / procedure The pretrained DNA transformer is adapted with a binary classifier using the study’s natural/artificial sequence labels. Individual claims	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2) Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` 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	MAGICS-LAB/DNABERT_2 README.md Original source ↗ README.md; checkpoint/access documentation and licence scope Version: f25bed9ee20db966dff39e5c1571249d04e36404 Retrieved: 2026-09-16T20:00:02.624261+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: `734a8cec5f667d74d421bf3b273ad7e256216109636da45aa7ceba21cd34de16` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Biological inputs 2-kbp natural or generated DNA sequences Individual claims	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4); Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1) Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Outputs Natural-versus-artificial sequence predictions Individual claims	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Methods/Training/Evaluation Datasets/Generated Sequence Discrimination (paragraph 1); Results/Model Safety (paragraph 1) Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Parameters An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources. Individual claims	MAGICS-LAB/DNABERT_2 README.md Original source ↗ Results/Pre-training; Methods/Model and Training/Tokenization; Methods/Model and Training/Architecture and Pre-training; Methods/Model and Training/Fine-tune GenomeOcean as Biosynthetic Gene Clusters Foundation Model (bgcFM); Methods/Training/Evaluation Datasets/Assembled Metagenome Datasets; Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity; Methods/Training/Evaluation Datasets/Biosynthetic Gene Cluster (BGC) Collection; Methods/Training/Evaluation Datasets/ZymoBiomics Microbial Community Standard Datasets; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: f25bed9ee20db966dff39e5c1571249d04e36404 Retrieved: 2026-09-16T20:00:02.624261+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.4.value` Source artifact SHA-256: `734a8cec5f667d74d421bf3b273ad7e256216109636da45aa7ceba21cd34de16` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Parameters An aggregate parameter total for this exact evaluated configuration is not established by the inspected sources. Individual claims	GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies Original source ↗ Results/Pre-training; Methods/Model and Training/Tokenization; Methods/Model and Training/Architecture and Pre-training; Methods/Model and Training/Fine-tune GenomeOcean as Biosynthetic Gene Clusters Foundation Model (bgcFM); Methods/Training/Evaluation Datasets/Assembled Metagenome Datasets; Methods/Training/Evaluation Datasets/Evaluation of Dataset Complexity; Methods/Training/Evaluation Datasets/Biosynthetic Gene Cluster (BGC) Collection; Methods/Training/Evaluation Datasets/ZymoBiomics Microbial Community Standard Datasets; inspected for aggregate parameter count (component sizes are not added without an exact configuration); README.md at pinned repository revision Shared locator for this statement’s cited sources; not a separate locator for each citation. Version: preprint archived 2025-02-05 Retrieved: 2026-09-16T10:33:55.224Z	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.4.value` Source artifact SHA-256: `3cc0df52522fccda23e3958f069c916b87ee50bb5c9a992fa37e25256546e145` 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

GenomeOcean: An Efficient Genome Foundation Model Trained on Large-Scale Metagenomic Assemblies · Original source · preprint archived 2025-02-05
MAGICS-LAB/DNABERT_2 README.md · Original source · f25bed9ee20db966dff39e5c1571249d04e36404
MAGICS-LAB/DNABERT_2 LICENSE · Original source · f25bed9ee20db966dff39e5c1571249d04e36404

Download this release

Technical metadata and extraction receipts

Stable ID: reported-model-7f1165b35f10e2

areas: microbes-communities
entity level: method
version: Not reported
reported name: DNABERT-2
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: genomeocean-2025; evidence-reported-base-dnabert2-readme-md; source locator: Results/Model Safety (paragraph 1); Results/Model Safety (paragraph 2) | README.md model description | Results/Model Safety (paragraph 1); Results/GenomeOcean Learns Protein-coding Principles from DNA Alone (paragraph 4); 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: DNABERT-2: Natural vs artificial microbial genome sequence

Suggest a correction