Together AI 在 ICML 2026 展示全栈前沿研究
原标题：Together AI at ICML 2026: frontier research across the full stack
AI 摘要
Together AI 在 ICML 2026 上发表了九篇论文，覆盖从智能体到 GPU 内核的全栈研究。其中，Aurora 论文提出的自适应投机解码已作为 ATLAS 投机器在生产环境中部署。研究还包括 DSGym 数据科学智能体评估框架、ThunderAgent 智能体工作负载优化、TTT-Discover 开放模型科学发现，以及 RARO 无验证器强化学习等方法。这些研究已集成到 Together 平台，并推动生产负载的改进。
正文节选
Nine papers is a lot to take in as a list. The better way to read them is by where they sit in the stack. Frontier AI is not built at a single layer. It is the product of research that runs from the agent down to the GPU kernel, and a gain at any one layer is wasted if the layers around it cannot keep up. This is at the core of how we work. From frontier agents at the top to kernels at the bottom, our research touches each one, and each layer feeds the next. The research becomes part of the Toge