RVDB 33-class viral classification by nearest-neighbour retrieval over frozen embeddings (Bůžková et al. 2026 Table 3)
Assign the organism label of the single nearest indexed training sequence to each test sequence, comparing pooling strategies across nine models.
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.
36 recorded evaluations, 36 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.
Results
Results are available, but no reviewed comparison panel is linked in this release.
All evaluations
36 evaluations · 36 results. Different protocols are not a single leaderboard.
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
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.
All 36 result rows with coverage, uncertainty and sources
Filter evaluations
Applied filters: All linked evaluations
Source checking is not independent reproduction. Release 2026-11-02-a7f03c9e55a5.
Methods and evaluation design
Procedure, tasks and evaluated configurations
Recorded evaluations
Each evaluation records what was tested and under which conditions.
- DNABERT-2, cls token only pooling, on the RVDB 33-class dataset
- DNABERT-2, max, all tokens pooling, on the RVDB 33-class dataset
- DNABERT-2, max, special tokens excluded pooling, on the RVDB 33-class dataset
- DNABERT-2, mean, all tokens pooling, on the RVDB 33-class dataset
- DNABERT-2, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- DNABERT-S, cls token only pooling, on the RVDB 33-class dataset
- DNABERT-S, max, all tokens pooling, on the RVDB 33-class dataset
- DNABERT-S, max, special tokens excluded pooling, on the RVDB 33-class dataset
- DNABERT-S, mean, all tokens pooling, on the RVDB 33-class dataset
- DNABERT-S, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- NTv1 (Multispecies 2.5B), cls token only pooling, on the RVDB 33-class dataset
- NTv1 (Multispecies 2.5B), max, all tokens pooling, on the RVDB 33-class dataset
Baseline coverage
Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.
0 of 2 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.
No execution recipe linked to this protocol. Recipe availability does not establish a completed evaluation.
- External evaluations
- 36
Literature evidence is not a Rewire measurement. Executed but unpublished runs and private review status are not included.
Null control
Proposed control: requires review
Training-set class prior where supervised fitting is permitted
Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.
This is a suggested selection rule, not a validated method or a measured score.
Conventional reference
Proposed control: requires review
Regularised classifier on simple permitted features, or protocol's conventional reference
Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.
This is a suggested selection rule, not a validated method or a measured score.
Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)
Coverage is derived from release 2026-11-02-a7f03c9e55a5. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.
Run instructions
No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.
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
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
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
Technical metadata and extraction receipts
Stable ID: gfm2-20261011-buzkova2026-protocol-acc-33
- areas
- dna-genomes
- contexts
- research
- protocol
- Generate one frozen embedding per sequence by the stated pooling, L2-normalise, index the training embeddings with FAISS IndexFlatIP, retrieve the nearest neighbour (k = 1) for each test sequence and take its label. No parameter update.
- version
- Table 3 (33-class dataset, maximum length 2,046 bp)
- metric
- accuracy
- metric direction
- higher
- unit
- fraction
- aggregation
- Share of the 4,950 test sequences labelled correctly
- metric implementation
- Accuracy over test sequences; macro F1, precision and recall reported separately in Table 7
- limitations
- Nearest-neighbour probe over labelled reference embeddings, not zero-shot classification: the index holds the training labels (Section 2.7).; Point estimates from one fixed split with a fixed random seed; 95% Wilson intervals are quoted in the text, not in the tables.; RVDB organism labels are not a controlled taxonomy; the 33-class refinement removed seven ambiguous labels found through the models' own confusions (Sections 3.2 and 4).; ICTV reconciliation found one further redundant label pair still inside the 33-class dataset (Section 3.2; Section 4, Table 11).
- source locator
- Table 3; Sections 2.7 and 3.1
Related records
- uses data: RVDB C-RVDB v31.0 primary dataset, 33 refined classes
- assessment: DNABERT-2, cls token only pooling, on the RVDB 33-class dataset
- assessment: DNABERT-2, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: DNABERT-2, max, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: DNABERT-2, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: DNABERT-2, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: DNABERT-S, cls token only pooling, on the RVDB 33-class dataset
- assessment: DNABERT-S, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: DNABERT-S, max, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: DNABERT-S, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: DNABERT-S, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv1 (Multispecies 2.5B), cls token only pooling, on the RVDB 33-class dataset
- assessment: NTv1 (Multispecies 2.5B), max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv1 (Multispecies 2.5B), max, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv1 (Multispecies 2.5B), mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv1 (Multispecies 2.5B), mean, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv2-250M, cls token only pooling, on the RVDB 33-class dataset
- assessment: NTv2-250M, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv2-250M, max, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv2-250M, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv2-250M, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv2-500M, cls token only pooling, on the RVDB 33-class dataset
- assessment: NTv2-500M, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv2-500M, max, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv2-500M, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv2-500M, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv2-50M, cls token only pooling, on the RVDB 33-class dataset
- assessment: NTv2-50M, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv2-50M, max, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv2-50M, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv2-50M, mean, special tokens excluded pooling, on the RVDB 33-class dataset
- assessment: NTv3-100M, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv3-100M, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv3-650M, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv3-650M, mean, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv3-8M, max, all tokens pooling, on the RVDB 33-class dataset
- assessment: NTv3-8M, mean, all tokens pooling, on the RVDB 33-class dataset