Genome Understanding Evaluation (GUE)
28 datasets/tasks in the native packaging; mapped here to five task families.
1 个 benchmark 数值协议 · 10 条 benchmark 全局结果（非 task 小计）D1 · 序列理解与调控
转录因子结合位点预测
Predict sequence-specific TF occupancy or binding sites.
TF binding-site prediction
任务定义
- 输入粒度
- sequence/window or base
- 生物学目标
- TF occupancy/site
- 输出表示
- probability or per-base score
- ML formulation
- classification/segmentation
- 计数说明
- NR
Benchmark coverage
7 个 benchmark 覆盖该任务
Nucleotide Transformer benchmark
Eighteen downstream tasks are benchmark instances, not eighteen unique task families.
67 个 benchmark 数值协议 · 821 条 benchmark 全局结果（非 task 小计）BEND
Human-genome tasks with biologically aligned splits; inspect each track for leakage controls.
7 个 benchmark 数值协议 · 105 条 benchmark 全局结果（非 task 小计）DART-Eval
Regulatory DNA evaluation across zero-shot, probing and fine-tuning regimes.
0 个 benchmark 数值协议 · 0 条 benchmark 全局结果（非 task 小计）Benchmarking DNA foundation models for genomic and genetic tasks
One-time benchmark study rather than a maintained benchmark suite.
6 个 benchmark 数值协议 · 8 条 benchmark 全局结果（非 task 小计）ENCODE-DREAM in vivo TF binding
Held-out cell types and chromosomes improve realism, but access is through Synapse.
0 个 benchmark 数值协议 · 0 条 benchmark 全局结果（非 task 小计）OmniGenBench
Meta-suite packaging RGB, BEACON, GUE, Genomic Benchmarks and PGB; native_task_count=5 suites, not the sum of their downstream tasks.
0 个 benchmark 数值协议 · 0 条 benchmark 全局结果（非 task 小计）评测协议
0 个口径单元
该任务已有 benchmark 映射，但当前没有可重建的数值协议。