Model type
Autoregressive DNA model; this record is the paper-specific evaluated configuration.
This configuration is evaluated for phage–bacterium discrimination and genome-wide prophage detection in LAMBDA.
Conceptual input–method–output guide. Check the procedure text and linked evaluation for fitted components, additional inputs and exact settings.
Autoregressive DNA model; this record is the paper-specific evaluated configuration.
Bacterial/phage DNA sequences and genome windows
Phage classification scores or candidate prophage regions
limited source coverage · Automated source review, 2026-09-16. All specifications and missing details
Release 2026-09-17-d277315f7d76 · 1 evaluation · 6 metric rows. Different protocols are not a single leaderboard.
| Metric and finding | Coverage and uncertainty | Evidence |
|---|---|---|
| EVO2: Genome-wide prophage detection Configuration: EVO2Protocol: Genome-wide prophage detection (Genome-wide prophage detection)Dataset: LAMBDA genome-wide prophage test Scan complete genomes with overlapping windows and postprocess predictions into prophage regions. Genome-wide metrics are computed at region level and macro-averaged across genomes. Filtered gLM pipelines include normalization, smoothing, clustering and size filtering; these are not standalone encoder scores. 80 bacterial genomes,47 species,four phyla;386 annotated prophage locations. Independent external evaluation · Evaluation metadata: needs review | ||
| 0.680 MCC Unit: unitless · Direction: higher | Uncertainty: not reported in legacy extract Scored: Not reported · Eligible: Not reported | source checkedLAMBDA: A Prophage Detection Benchmark for Genomic Language Models; LAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table 5, EVO2 row, MCC column Source checking is not independent reproduction. |
| 0.012 FPR Unit: fraction · Direction: lower | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table 5., row EVO2†, column FPR; XML row8 column5 Source checking is not independent reproduction. |
| 0.694 Precision Unit: fraction · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table 5., row EVO2†, column Precision; XML row8 column2 Source checking is not independent reproduction. |
| 0.665 F1 Unit: fraction · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table 5., row EVO2†, column F1; XML row8 column6 Source checking is not independent reproduction. |
| 0.988 Specificity Unit: fraction · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table 5., row EVO2†, column Specificity; XML row8 column4 Source checking is not independent reproduction. |
| 0.742 Recall Unit: fraction · Direction: higher | Uncertainty: unreported Scored: Not reported · Eligible: Not reported | source checkedLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table 5., row EVO2†, column Recall; XML row8 column3 Source checking is not independent reproduction. |
The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used.
Evo 2 uses the StripedHyena 2 architecture at single-nucleotide resolution. The official family documentation describes autoregressive training on 8.8 trillion OpenGenome2 tokens and context lengths up to one million bases; an exact evaluated checkpoint remains a separate identity.
The linked evaluation record identifies EVO2: Genome-wide prophage detection. Its dataset, split, adaptation and evidence origin remain attached to the reported results.
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-aa763db2cfdeffExplanatory 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.
| Property | Description and evidence |
|---|---|
| Model type | Autoregressive DNA model; this record is the paper-specific evaluated configuration.SourcesArcInstitute/evo2 README.md · README.md model description |
| Architecture / procedure | The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used.SourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3) |
| Biological inputs | Bacterial/phage DNA sequences and genome windowsSourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Materials and Methods/LAMBDA Benchmark Dataset Construction./Segment Subsampling. (paragraph 1); Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Sources. (paragraph 1) |
| Outputs | Phage classification scores or candidate prophage regionsSourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Results/Genome-Wide Prophage Detection./Scanning for Novel, Unannotated Prophage Regions. (paragraph 2); Materials and Methods/LAMBDA Benchmark Dataset Construction./Prophage Signal Extraction Algorithm. (paragraph 1) |
| Parameters | 7B parameters in the benchmark model inventory.SourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table1; EVO2 row |
| Known versions / configuration | EVO2 7B as identified by Table 1; exact checkpoint hash is not supplied in that inventory.SourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Results/Genome-Wide Prophage Detection./Probing EVO2 Latent Space with Sparse Autoencoders. (paragraph 1); Table T2 (paragraph 1) |
| Training data / fitting | OpenGenome2 multispecies pretraining, as identified in Table 1; LAMBDA trains separate linear or three-layer neural probes on frozen embeddings.SourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Results/Evaluating the Power of Pretrained Sequence Representations. (paragraph 2); Materials and Methods/LAMBDA Benchmark Dataset Construction./Comparison with Traditional and Protein-based Models. (paragraph 1) |
| Context limits | The source lists a 1M-base model limit; the actual LAMBDA experiments evaluate 2-kb,4-kb and 8-kb windows.SourcesLAMBDA: A Prophage Detection Benchmark for Genomic Language Models · Table1 EVO2 row; Tables2–3 window lengths |
| Access | Official upstream implementation and usage documentation: https://github.com/ArcInstitute/evo2/blob/53f195997257c56c00e5ef8d33a54f5baad143a6/README.md. This pinned documentation revision is not automatically the evaluated weight revision.SourcesArcInstitute/evo2 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).SourcesArcInstitute/evo2 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 sourcesSourcesArcInstitute/evo2 README.md · README.md; checkpoint/access documentation and licence scope |
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.
20 evidence rows matching the loaded filters
| 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 | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Diagram steps ["Bacterial/phage DNA sequences and genome windows","EVO2","Phage classification scores or candidate prophage regions"] Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Diagram title Evaluated procedure (conceptual) Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Model type Autoregressive DNA model; this record is the paper-specific evaluated configuration. Individual claims | ArcInstitute/evo2 README.md README.md model description Version: 53f195997257c56c00e5ef8d33a54f5baad143a6 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Architecture / procedure The benchmark separates embedding probes, fine-tuning, diagnostic tests and genome-wide scans; the associated result identifies which procedure is used. Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Abstract (paragraph 1); Results/Genome-Wide Prophage Detection./Extracting Prophage Signal from Raw Model Predictions. (paragraph 3) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| 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 | ArcInstitute/evo2 README.md README.md; checkpoint/access documentation and licence scope Version: 53f195997257c56c00e5ef8d33a54f5baad143a6 | unreported automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Biological inputs Bacterial/phage DNA sequences and genome windows Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Materials and Methods/LAMBDA Benchmark Dataset Construction./Segment Subsampling. (paragraph 1); Materials and Methods/LAMBDA Benchmark Dataset Construction./Data Sources. (paragraph 1) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Outputs Phage classification scores or candidate prophage regions Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Results/Genome-Wide Prophage Detection./Scanning for Novel, Unannotated Prophage Regions. (paragraph 2); Materials and Methods/LAMBDA Benchmark Dataset Construction./Prophage Signal Extraction Algorithm. (paragraph 1) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Parameters 7B parameters in the benchmark model inventory. Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Table1; EVO2 row Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
| Known versions / configuration EVO2 7B as identified by Table 1; exact checkpoint hash is not supplied in that inventory. Individual claims | LAMBDA: A Prophage Detection Benchmark for Genomic Language Models Results/Genome-Wide Prophage Detection./Probing EVO2 Latent Space with Sparse Autoencoders. (paragraph 1); Table T2 (paragraph 1) Version: PMC13041943.1 | source checked automated source review · 2026-09-16 Audit detailsPrimary 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: Source artifact SHA-256: Hash scope: Hash scope not separately documented; inspect source record |
Release 2026-09-17-d277315f7d76 · Record review: needs review
Stable ID: reported-model-aa763db2cfdeff