Configuration

NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)

NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN, a configuration of Nucleotide Transformer: 650M parameters; context 1 Mb; Single-base tokenizer; frozen weights; Max, all tokens pooling; L2-normalised embeddings, FAISS IndexFlatIP, k = 1.

3 evaluations · 9 results 2 open disagreements between studies involve its model

Overview

An explanatory profile has not yet been reviewed for this record.

Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.

Evaluations and results

3 evaluations · 9 results. Different protocols are not a single leaderboard.

Sorted by Accuracy (33-class RVDB dataset, 1-NN over frozen embeddings) (higher is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.

Tested configuration	Protocol and dataset	Accuracy (33-class RVDB dataset, 1-NN over frozen embeddings)	Accuracy (33-class RVDB dataset)	Accuracy (40-class RVDB dataset)	Macro F1 (33-class RVDB dataset)	Macro F1 (40-class RVDB dataset)	Precision (33-class RVDB dataset, macro-averaged over classes)	Precision (40-class RVDB dataset, macro-averaged over classes)	Recall (33-class RVDB dataset, macro-averaged over classes)	Recall (40-class RVDB dataset, macro-averaged over classes)
NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)Evaluation	RVDB 33-class viral classification by nearest-neighbour retrieval over frozen embeddings (Bůžková et al. 2026 Table 3), RVDB C-RVDB v31.0 primary dataset, 33 refined classes	⋆0.9265	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)Evaluation	RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7), RVDB C-RVDB v31.0 primary dataset, 33 refined classes	Not reported	0.926	Not reported	0.925	Not reported	0.926	Not reported	0.926	Not reported
NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)Evaluation	RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7), RVDB C-RVDB v31.0 primary dataset, 40 sampled classes	Not reported	Not reported	0.875	Not reported	0.874	Not reported	0.874	Not reported	0.875

All 9 result rows with coverage, uncertainty and sources

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: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB 33-class viral classification by nearest-neighbour retrieval over frozen embeddings (Bůžková et al. 2026 Table 3) Dataset: RVDB C-RVDB v31.0 primary dataset, 33 refined classes	⋆0.9265 Accuracy (33-class RVDB dataset, 1-NN over frozen embeddings) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M, max, all tokens pooling, on the RVDB 33-class dataset gfm2-20261011-buzkova2026-protocol-acc-33 Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 3, column 'NTv3-650M', row 'Max, all tokens'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 33 refined classes	0.926 Accuracy (33-class RVDB dataset) fraction · higher Uncertainty: 95% CI 0.919 to 0.933. Wilson binomial interval quoted in Section 3.6, paragraph 1; method stated in Section 2.9 Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 33-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '33 classes / NTv3-650M (max)', column 'Accuracy'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 33 refined classes	0.925 Macro F1 (33-class RVDB dataset) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 33-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '33 classes / NTv3-650M (max)', column 'F1'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 33 refined classes	0.926 Precision (33-class RVDB dataset, macro-averaged over classes) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 33-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '33 classes / NTv3-650M (max)', column 'Precision'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 33 refined classes	0.926 Recall (33-class RVDB dataset, macro-averaged over classes) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 33-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '33 classes / NTv3-650M (max)', column 'Recall'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 40 sampled classes	0.875 Accuracy (40-class RVDB dataset) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 40-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '40 classes / NTv3-650M (max)', column 'Accuracy'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 40 sampled classes	0.874 Macro F1 (40-class RVDB dataset) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 40-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '40 classes / NTv3-650M (max)', column 'F1'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 40 sampled classes	0.874 Precision (40-class RVDB dataset, macro-averaged over classes) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 40-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '40 classes / NTv3-650M (max)', column 'Precision'
Configuration: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026)	Protocol: RVDB viral classification: best frozen-embedding configuration against BLAST (Bůžková et al. 2026 Table 7) Dataset: RVDB C-RVDB v31.0 primary dataset, 40 sampled classes	0.875 Recall (40-class RVDB dataset, macro-averaged over classes) fraction · higher Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Methods, coverage and source NTv3-650M (max) on the RVDB 40-class dataset (BLAST comparison) gfm2-20261011-buzkova2026-protocol-blast-compare Aggregation: Not reported Analyzing Genomic Foundation Models for Viral Sequence Identification · Table 7, row '40 classes / NTv3-650M (max)', column 'Recall'

Source checking is not independent reproduction. Release 2026-11-02-a7f03c9e55a5.

Use this configuration

How it works, versions and access

Additional methodological details, versions and access terms have not yet been extracted into a structured profile for this configuration. Consult the linked sources and evaluation records below.

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.

0 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-11-02-a7f03c9e55a5
Property and statement	Original source and location	Review and provenance

No evidence rows match these filters. Choose another scope or clear the search.

How to read the evidence

Sources and history

Release 2026-11-02-a7f03c9e55a5 · Record review: source checked

1 source record and release history

Analyzing Genomic Foundation Models for Viral Sequence Identification · Original source · bioRxiv v1, posted 2026-09-28

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: gfm2-20261011-buzkova2026-config-ntv3-650m-max-all

areas: dna-genomes
contexts: research
method types: foundation_model
reported name: NTv3-650M
foundation model eligible: true
source locator: Table 2 row 'NTv3-650M'; Table 3 column 'NTv3-650M', row 'Max, all tokens'
parameters: 650M parameters; context 1 Mb; Single-base tokenizer; frozen weights; Max, all tokens pooling; L2-normalised embeddings, FAISS IndexFlatIP, k = 1
missing metadata: version: reason: unreported; note: Official Hugging Face repository named but no revision is printed (Section 2.4)
training setting: Frozen weights, no parameter update; nearest-neighbour probe over labelled reference embeddings
provenance note: NTv3-650M, frozen embeddings, max, all tokens pooling, 1-NN (Bůžková et al. 2026), as evaluated in the cited comparison.

Related records

configuration of: Nucleotide Transformer
system: NTv3-650M, max, all tokens pooling, on the RVDB 33-class dataset
system: NTv3-650M (max) on the RVDB 33-class dataset (BLAST comparison)
system: NTv3-650M (max) on the RVDB 40-class dataset (BLAST comparison)

Suggest a correction