Research map / Architectures and efficiency / Attention and transformers
HyenaDNA: Long-Range Genomic Sequence Modeling at Single Nucleotide Resolution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| HyenaDNA (single-nucleotide Hyena) vs CNN baseline | genomic classifiers · GenomicBenchmarks Human Enhancers Ensembl | accuracy: 68.9 -> 89.2 | Helped |
| 1.6M HyenaDNA vs 2.5B Nucleotide Transformer | NT benchmarks · H3K4me3 | MCC: 42.1 -> 61.2 | Helped |
| 1.6M HyenaDNA vs 2.5B Nucleotide Transformer | NT benchmarks · Splice acceptor | F1: 99.0 -> 96.6 | Hurt |
| pretraining on human genome (ablation) | HyenaDNA · NT H3K4me3 | MCC: 40.2 -> 61.2 | Helped |
| pretraining on human genome (ablation) | HyenaDNA · GenomicBenchmarks | accuracy: 93.3 -> 96.6 | Helped |
| use single-nucleotide tokens instead of 6-mer tokenizer (ablation) | HyenaDNA · GenomicBenchmarks | accuracy: 83.5 -> 93.3 | Helped |
| use causal instead of bidirectional Hyena (ablation) | HyenaDNA from scratch · GenomicBenchmarks | accuracy: 3.8 accuracy points | Helped |
| replace attention with Hyena operator (same 1.6M size, pretrained) (ablation) | GPT · NT H3K4me3 | MCC: 28.3 -> 61.2 | Helped |
| add sequence length warm-up schedule (ablation) | HyenaDNA at 450k · species classification | accuracy: +7.5 accuracy points | Helped |
| increase context length from 1k to 1M (ablation) | HyenaDNA · species classification | accuracy: 61.1 -> 99.5 | Helped |
| add more k-shot demonstrations in context (ablation) | HyenaDNA · GenomicBenchmarks | accuracy | No clear effect |
| add more soft tuneable prompt tokens (ablation) | frozen HyenaDNA · GenomicBenchmarks | accuracy | Helped |
Closest papers by meaning
- Hyperbolic Genome Embeddings (ICLR 2025 · similarity 0.87)
- DNABERT-2: Efficient Foundation Model and Benchmark For Multi-Species Genomes (ICLR 2024 · similarity 0.86)
- JanusDNA: A Powerful Bi-directional Hybrid DNA Foundation Model (NeurIPS 2025 · similarity 0.86)
- BEND: Benchmarking DNA Language Models on Biologically Meaningful Tasks (ICLR 2024 · similarity 0.86)
- Context-Aware Regularization with Markovian Integration for Attention-Based Nucleotide Analysis (NeurIPS 2025 · similarity 0.86)
- What DNA Foundation Models Learn Beyond Sequence Composition (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.