Everyone can now discover new SOTA in science with a few hundred bucks! Test-Time Training + open m...
TL;DR · AI 摘要
斯坦福AI实验室提出Test-Time Training结合开源模型，在数学、算法等领域以低成本实现超越闭源大模型的科学发现能力。
核心要点
- Test-Time Training让AI在解决具体问题时持续学习，优于仅靠提示工程的方法
- 开源模型+Test-Time Training在科学发现任务上胜过Gemini、GPT-5等闭源前沿模型
- 该方法使个人能以数百美元成本参与前沿科学探索
Test-Time Training + open model > prompt engineering + closed frontier model (Gemini, GPT-5), for discovery problems in Mathematics, Kernel Engineering, Algorithms and Biology.
https://t.co/IIN67Abc9H" / X
Stanford AI Lab on X: "Everyone can now discover new SOTA in science with a few hundred bucks! Test-Time Training + open model > prompt engineering + closed frontier model (Gemini, GPT-5), for discovery problems in Mathematics, Kernel Engineering, Algorithms and Biology. https://t.co/IIN67Abc9H" / X
Don’t miss what’s happening
Everyone can now discover new SOTA in science with a few hundred bucks! Test-Time Training + open model > prompt engineering + closed frontier model (Gemini, GPT-5), for discovery problems in Mathematics, Kernel Engineering, Algorithms and Biology. https://test-time-training.github.io/discover.pdf
Quote
Mert Yuksekgonul
@mertyuksekgonul
·
Jan 22
How to get AI to make discoveries on open scientific problems? Most methods just improve the prompt with more attempts. But the AI itself doesn't improve. With test-time training, AI can continue to learn on the problem it’s trying to solve: http://test-time-training.github.io/discover.pdf
13
36
217
199
Read 13 replies