核酸结果 概览 Benchmark 任务 方法学 下载 快照 2026-08-05 2026-08-06 Tier B · benchmark_suite
有公开数值结果 Genome Understanding Evaluation (GUE) 28 datasets/tasks in the native packaging; mapped here to five task families.
B01 community_protocol not_applicable
结果覆盖
10 条结果， 1 个协议
主来源覆盖 primary_sources_screened
相关工作覆盖 priority_search_complete_with_logged_api_limits
权威性准备度 verified_for_internal_use
文献截止 2026-08-06 Track 数值覆盖 来源闭环 阻塞原因 B01-TR01Genome Understanding Evaluation (GUE) — Genomic element classification numeric_results_available10 results closed_with_numeric_evidence NR B01-TR02Genome Understanding Evaluation (GUE) — Promoter recognition numeric_results_available0 results closed_no_eligible_public_numeric_result related_work_or_configuration_fields_incomplete B01-TR03Genome Understanding Evaluation (GUE) — Enhancer recognition numeric_results_available0 results closed_no_eligible_public_numeric_result related_work_or_configuration_fields_incomplete B01-TR04Genome Understanding Evaluation (GUE) — TF binding-site prediction numeric_results_available0 results closed_no_eligible_public_numeric_result related_work_or_configuration_fields_incomplete B01-TR05Genome Understanding Evaluation (GUE) — ncRNA family classification numeric_results_available0 results closed_no_eligible_public_numeric_result related_work_or_configuration_fields_incomplete
协议级结果
先选口径，再比较分数。 每个卡片仅展示同一 protocol + metric fingerprint 的四类声明。
28 GUE datasets across seven task categories · official train/validation/test splits · paper fine-tuning protocol; LoRA used for Nucleotide Transformer models
GUE average score mean across 28 datasets with task-specific MCC/F1 · 10 条结果 官方 baseline
DNABERT-2 0.668 GUE average score · fraction 证据 ↗ 原论文表内最佳
DNABERT-2 + GUE further pre-training 0.6777 GUE average score · fraction 证据 ↗ 官方榜单 / 挑战领先者
未建立该类声明 当前快照没有符合口径的结果。 严格跨工作 SOTA
未建立该类声明 不会用表内最佳值代替跨工作 SOTA。 指纹: PMF-v1-3D5414ACEBD82B2388042A87