Researchers have developed Thoth, a scientific reasoning model designed to generate biologically sound and executable experimental protocols. Unlike previous models that often produced protocols with missing steps or incorrect parameters, Thoth focuses on structured reasoning to ensure logical order and semantic accuracy. The model was trained using a novel "Sketch-and-Fill" paradigm and a "SCORE" reward mechanism that prioritizes experimental feasibility over mere textual similarity. In evaluations, Thoth demonstrated superior performance compared to several large language models, including GPT-4o, on tasks requiring precise scientific reasoning and protocol generation. AI
影响 Enables AI to move beyond answering questions to generating executable scientific experiments, potentially accelerating life science research.
排序理由 The cluster describes a new scientific reasoning model and its associated paper presented at a major AI conference. [lever_c_demoted from research: ic=1 ai=1.0]
- ChatGPT-4o
- DeepSeek-V3
- Fudan University
- GPT-4o
- ICLR 2026
- Qwen3-8B
- SciRecipe
- Shanghai Artificial Intelligence Laboratory
- Shanghai Jiao Tong University
- Thoth
AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →
