GENEB: Genomic Embedding Benchmark Tasks for DNA Sequence Encoders

GENEB provides task-level sequence classification data for benchmarking DNA sequence encoders. The dataset contains train and test partitions for 100 tasks across 13 functional categories, intended for evaluation with precomputed representations and logistic regression. It was authored by DeepPavlov and last updated in June 2026.

Use Cases

• Benchmarking DNA sequence encoder performance based on 100 classification tasks.
• Evaluating precomputed genomic embeddings without fine-tuning on task labels.
• Comparing model capabilities across 13 functional categories of genomic tasks.

Strengths

• Contains 100 distinct classification tasks for multi-faceted evaluation.
• Tasks are organized into 13 functional categories for structured analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset scale are unknown, which may limit suitability assessment.

Provenance

Source

DeepPavlov

Freshness

Last updated 2026-06-04 14:51:59; freshness should be verified.