Paving the way for agents in biology(為 AI Agent 鋪平生物學之路)
Paving the way for agents in biology(為 AI Agent 鋪平生物學之路)
來源: Anthropic 原文日期: 2026-06-08
中文摘要
Anthropic 研究員 Laura Luebbert 指出,生物學 AI Agent 發展的最大瓶頸不是模型推理能力,而是生物數據基礎設施對 agent 極不友好。研究團隊讓 Claude、Biomni OSS、Edison Analysis 和 GPT 等 agent 從 NCBI Virus 病毒数据库中擷取序列數據,結果最強模型的準確率也不一致(16.9%–91.3%),且同樣問題三次跑出的結果差異巨大。團隊開發了「gget virus」確定性檢索層後,準確率一舉提升至近 100%。
關鍵洞察
原文關鍵句(英文保留)
> "Even the strongest models did not consistently achieve the level of accuracy required for reliable dataset construction. But accuracy rose to nearly 100% once she and her team added gget virus, a deterministic retrieval layer."
> "Software infrastructure was basically made for the needs of cars (agents): paved roads, clear lanes, standardized signals. Using AI agents to navigate biological data infrastructure is like driving through an old city that was designed before cars."
> "The bottleneck for biological agents is not only reasoning but the absence of widespread deterministic execution layers for querying biological data."